ANALYTICAL BIOCHEMISTRY 251, 89–97 (1997)ARTICLE NO. AB972250

Milk Oligosaccharide Profiles by Reversed-Phase HPLCof Their Perbenzoylated Derivatives1

Prasoon Chaturvedi,*,† Christopher D. Warren,* Guillermo M. Ruiz-Palacios,‡Larry K. Pickering,§ and David S. Newburg*,†*Eunice Kennedy Shriver Center for Mental Retardation, Waltham, Massachusetts 02254; †Department ofNeurology, Harvard Medical School, Boston, Massachusetts 02115; ‡Institute Nacional de Nutricion,Mexico City, Mexico; and §Eastern Virginia Medical School, Norfolk, Virginia 23510

Received February 21, 1997

and pathogens involved in diseases of infants (2). TheseHuman milk is rich in oligosaccharides, some of oligosaccharides are the third largest solid component in

which inhibit toxins and pathogens involved in dis- milk after lactose and triglycerides and are believed toeases of infants. To investigate qualitative and quanti- be synthesized by the same glycosyltransferases that aretative individual variation of human milk oligosaccha- involved in the biosynthesis of glycoproteins and otherrides, a sensitive method for routine identification and glycoconjugates. As a consequence, milk oligosaccharidesquantification of intact milk oligosaccharides was de- can have terminal carbohydrate structures similar toveloped and applied to milk samples from 50 donors. those present on glycoconjugates, some of which act asThe isolated, reduced neutral oligosaccharide frac- cell surface receptors for pathogens. Thus, some of thesetions were perbenzoylated, resolved by reversed- oligosaccharides could act as analogs or homologs tophase HPLC, and detected at 229 nm. This method re- those receptors, thereby providing protection to the nurs-solves most structural isomers and does not require ing infant against enteric pathogens. Recent studies onstringent removal of lactose. Peaks were detected at

complex sugar structures (3) have revealed the existencethe low nanogram (pmol) level and peak areas wereof a far greater number of oligosaccharides than whatlinear from 1 to 1000mg for a standard oligosaccharide.was hitherto known, further widening the interest inOligosaccharide samples equivalent to 1 ml of humantheir possible biological roles. Although there have beenmilk give optimum chromatographic separation anda number of studies on the qualitative aspects of milkresolution. The method gives quantitative results com-oligosaccharides, few reports have appeared on theirparable to those obtained with classic total sugar anal-quantification, especially with regard to quantitativeyses, and has an average coefficient of variation of 13%.variations of individual intact oligosaccharides from dif-The 12 major peaks in human milk coeluted with au-ferent donors. In a recent report by Miller et al. (4), indi-thentic oligosaccharide standards ranging from tri- to

octasaccharides, and their identities were confirmed vidual and temporal variations were observed in the con-by mass spectrometry. Significant individual variation tent of lactose and of monosaccharide componentsexists in oligosaccharide profiles; almost 70% of sam- derived by hydrolysis of human milk oligosaccharides.ples contained 2 *-fucosyllactose and lacto-N-fucopen- More recently, Thurl et al. (5), using the milk of a singletaose I as the major oligosaccharides; for the remain- donor, have described a method for quantification of milkder, the major oligosaccharides were 3-fucosyllactose oligosaccharides using high-pH anion-exchange chroma-and lacto-N-fucopentaose-II or lacto-N-fucopentaose- tography.III. This method can be used to investigate the extent Our earlier studies have demonstrated individualand biological significance of oligosaccharide varia- variations in milk fucosyltransferase activities as welltion in human milk. q 1997 Academic Press as changes in activity during lactation. Variations ob-

served in fucosyltransferase activities in milk could bereflective of analogous variations of their activity inthe cells of the mammary gland during oligosaccharide

Human milk contains a complex and wide array of biosynthesis. Such variations would be expected to leadoligosaccharides (1), some of which protect against toxins to qualitative and quantitative differences in the oligo-

saccharides of milk from different donors, as well asdifferences in milk at different stages of lactation. A1 Support by HD13021.

890003-2697/97 $25.00Copyright q 1997 by Academic PressAll rights of reproduction in any form reserved.

AID AB 2250 / 6m3e$$$221 07-24-97 14:49:02 abal

CHATURVEDI ET AL.90

TABLE 1

Structures of Human Milk Oligosaccharides Identified in This Study

Abbreviation Trivial name Structure

Lac Lactose Gal b(1 r 4) Glc2 *-FucLac 2 *-Fucosyllactose Fuc a(1 r 2) Gal b(1 r 4) Glc3-FucLac 3-Fucosyllactose Gal b(1 r 4) x

GlcFuc a(1 r 3) z

LDFT Lactodifucotetraose Fuc a(1 r 2) Gal b(1 r 4) x

GlcFuc a(1 r 3) z

LNT Lacto-N-tetraose Gal b(1 r 3) GlcNAc b(1 r 3) Gal b(1 r 4) GlcLNneoT Lacto-N-neotetraose Gal b(1 r 4) GlcNAc b(1 r 3) Gal b(1 r 4) GlcLNF-I Lacto-N-fucopentaose-I Fuc a(1 r 2) Gal b(1 r 3) GlcNAc b(1 r 3) Gal b(1 r 4) GlcLNF-II Lacto-N-fucopentaose-II Gal b(1 r 3) x

GlcNAc b(1 r 3) Gal b(1 r 4) GlcFuc a(1 r 4) z

LNF-III Lacto-N-fucopentaose-III Gal b(1 r 4) x

GlcNAc b(1 r 3) Gal b(1 r 4) GlcFuc a(1 r 3) z

LDFH-I Lacto-N-difucohexaose-I Fuc a(1 r 2) Gal b(1 r 3) x

GlcNAc b(1 r 3) Gal b(1 r 4) GlcFuc a(1 r 4) z

LDFH-II Lacto-N-difucohexaose-II Gal b(1 r 3) x

GlcNAc b(1 r 3) Gal b(1 r 4) x

Fuc a(1 r 4) z GlcFuc a(1 r 3) z

MFLNH-III Monofucosyllacto-N-hexaose-III Fuc a(1 r 3) x

GlcNAc b(1 r 6) x

Gal b(1 r 4) z Gal b(1 r 4) GlcGal b(1 r 3) GlcNAc b(1 r 3) z

DFLNHa Difucosyllacto-N-hexaose-a Fuc a(1 r 3) x

GlcNAc b(1 r 6) x

Gal b(1 r 4) z Gal b(1 r 4) GlcFuc a(1 r 2) Gal b(1 r 3) GlcNAc b(1 r 3) z

further source of differences in oligosaccharide levels random human milk samples and have assessed thevariability of oligosaccharide content among these sam-and structures could be the activity of a-L-fucosidase,

which has been shown to be active in milk (6). This ples. The sensitivity and linearity of the method wasassessed using commercially available oligosaccharidefucosidase activity also showed variations among indi-

vidual donors and varied over the course of lactation standards.within individuals. In addition, activities of other glyco-

MATERIALS AND METHODSsyltransferases such as b-N-acetylglucosaminyltrans-ferase might also vary, resulting in differences in oligo- Human milk samples were from healthy lactating

mothers 30 to 60 days postpartum living in Mexico Citysaccharide profiles among individual donors and fromdifferent stages of lactation. (San Pedro Martir) whose infants were healthy for 2

weeks prior and subsequent to the donation of milk.With mounting evidence supporting a role for milkoligosaccharides in protection of infants against enteric The samples were immediately placed on ice, and were

frozen within 1 h of collection and stored at 0807Cpathogens (2), variation in milk oligosaccharides fromdifferent donors and the relationship between such until used. Milk oligosaccharides were identified with

standards purchased from Oxford GlycoSystems (Bed-variation and differential susceptibility of infants tobacterial pathogens are important issues. As a first ford, MA). Sodium borohydride (NaBH4) and dimethyl-

aminopyridine were purchased from Aldrich (Milwau-step, the present study was aimed at developing aquick, reliable, and sensitive method for routine identi- kee, WI). Benzoic anhydride was from Sigma (St. Louis,

MO). Acetonitrile (HPLC grade) was obtained fromfication and quantification of intact milk oligosaccha-rides. We have used reversed-phase HPLC of perbenzo- Fisher (Pittsburgh, PA). Double distilled deionized wa-

ter was used for all HPLC analyses. All other reagentsylated derivatives of intact milk oligosaccharides toidentify and quantify 12 major milk oligosaccharides and solvents were ACS analytical grade or better and

were obtained from commercial sources.(for structures and abbreviations, see Table 1) from 50

AID AB 2250 / 6m3e$$$222 07-24-97 14:49:02 abal

CHROMATOGRAPHIC PROFILING OF MILK OLIGOSACCHARIDES 91

Isolation of oligosaccharides. Oligosaccharides were zoic anhydride (50 mg/ml) and 4-dimethylaminopyri-dine (25 mg/ml) in dry pyridine] was added (pyridineisolated from milk using the scheme shown in Fig. 1.

Milk samples (0.5 ml) were thawed immediately before was predried over three sequential treatments of 4Amolecular sieves that had been freshly baked in anuse, diluted with an equal quantity of water, and cen-

trifuged at 4000g for 45 min at 47C. The viscous, upper oven at 4507C overnight and cooled in a desiccator be-fore use). After mixing, incubation was for 16 h at 377C.cream layer, consisting primarily of fats and other lip-

ids, was removed by filtering through a glass wool plug Water (4.5 ml) was added to each sample and the re-sulting solution passed twice over a C-18 Bond-Elutin a Pasteur pipet or by careful pipetting from the

lower, aqueous layer. To the filtrate, or aqueous layer, column (3 ml, 0.5 mg; Varian, Sunnyvale, CA) fittedonto a vacuum manifold, with adequate vacuum towas added 2 ml of ethanol (to give 66.7% ethanol), and

the sample was kept overnight at 47C. The precipitate, achieve a flow rate of 0.5 ml/min. The C-18 columnshad previously been wetted with HPLC acetonitrile (5consisting predominantly of proteins and lactose, was

removed by centrifugation at 4000g for 15 min at 47C. ml) and equilibrated with 5 ml of 1% pyridine in water.After washing the column with 5 ml of 10% pyridineThe clear supernatant, consisting primarily of oligosac-

charides and residual lactose, was transferred to a and 5 ml of HPLC water, the perbenzoylated oligosac-charides were eluted with 5 ml of HPLC acetonitrile,screw-cap tube, dried by passage of nitrogen, lyophi-

lized, and weighed. and the eluate was dried under N2 and taken up in 100ml of HPLC acetonitrile.Reduction and perbenzoylation of milk oligosaccha-

rides was carried out by a modification of a method The resulting perbenzoylated oligosaccharides wereresolved by reversed-phase HPLC (Rainin C-8 column,described by Daniel (7) for the HPLC analysis of high-

mannose-type oligosaccharides. The lyophilized resi- 3 m; 4.6 mm 1 10 cm) at 1 ml/min with a 15-min lineargradient from acetonitrile:water (4:1) to 100% acetoni-due containing milk oligosaccharides was treated with

0.5 ml of a freshly made aqueous solution of sodium trile, with holding under the final conditions for anadditional 10 min. The perbenzoylated oligosaccha-borohydride (10 mg/ml). After vortex mixing, the reduc-

tion mixture was kept overnight at room temperature, rides were detected at 229 nm and their peaks inte-grated on a Macintosh computer with Rainin Dynamaxwhereupon the reaction was terminated by the drop-

wise addition of 1 M acetic acid until evolution of H2 software (Emeryville, CA).gas stopped (five to six drops). To remove sodium cat-ions, the resulting solution of reduced milk oligosaccha- RESULTS AND DISCUSSIONrides was applied to a Pasteur pipet column (5 1 0.5 The isolation scheme shown in Fig. 1 separates hu-cm) of AG50W-X8 cation-exchange resin [pyridinium man milk oligosaccharides into acidic and neutral spe-form, previously prepared by stirring AG50 resin, hy- cies. Although the acidic oligosaccharides are of poten-drogen form (Bio-Rad, Hercules, CA), with 10% excess tial importance, our interest has been on biologicallyaqueous pyridine solution, followed by washing with active components of the neutral oligosaccharide frac-an excess of deionized water]. The sample tube and tion, and therefore we have initially directed our at-column were washed with water in three 1-ml aliquots tention to their analysis. That notwithstanding, pre-and the combined flow-through and wash eluates dried liminary experiments indicate that with suitableby passage of nitrogen. Methanol (0.5 ml) was added modification this procedure may also be applied to theand evaporated thrice under nitrogen to remove boric acidic oligosaccharides.acid as methyl borate. For separation into acidic (sialic

Quantification. Figure 2A illustrates that the de-acid-containing) and neutral species, the reduced milktector response for perbenzoylated lactose is linear overoligosaccharides were dissolved in 0.5 ml water andthe exceptionally wide range tested (1 mg to 1 mg).applied to a column (5 1 0.5 cm) of AG1-X8 anion-Furthermore, a useful response was apparent at õ10exchange resin (Bio-Rad, acetate form, prewashed withpmol (õ10 ng); even at 6 pmol of lactose, the signal to10 ml of water). The sample tube and column werenoise ratio (S:N) is 30:1. Likewise, LNF I,2 representa-washed with water in three 1-ml aliquots and the flow-tive of the higher oligosaccharides, was perbenzoylatedthrough and wash eluates collected in a weighed testand analyzed by HPLC. Figure 2B demonstrates thattube. After reducing the volume of the solution to 1 mlthe response of perbenzoylated LNF-I was likewise lin-by passage of nitrogen, the tube contents were lyophi-ear over the range tested (5 to 100 mg); furthermore,lized and weighed to yield the neutral milk oligosaccha-

rides. Washing of the AG1 resin column with 3 ml of0.5 M pyridinium acetate, followed by reduction of the 2 Abbreviations used: DP, degree of polymerization (i.e., number

of monosaccharide components per molecule); DFLNHa, difucosyl-solution volume to 1 ml and lyophilization, yielded thelacto-N-hexaose-a; 2*-FucLac, 2 *-fucosyllactose; 3-FucLac, 3-fucosyl-acidic milk oligosaccharides, which were reserved forlactose; GlcNAc, N-acetylglucosamine; LDFH-I, lacto-N-difucohex-future use. Neutral milk oligosaccharides (500 mg) were aose-I; LNF-I, II, III, lacto-N-fucopentaoses-I, II, III; LNT, lacto-N-

dried in vacuo over phosphorus pentoxide for 4 to 6 h. tetraose; LNneoT, lacto-N-neotetraose; MFLNH-III, monofucosyl-lacto-N-hexanose-III.To each sample, 0.5 ml perbenzoylation reagent [ben-

AID AB 2250 / 6m3e$$$222 07-24-97 14:49:02 abal

CHATURVEDI ET AL.92

II and MFLNH-III, both of which contained a peakthat is probably an artifact of this technique, but mayrepresent a small saccharide contaminant. This HPLCmethod could resolve even those milk oligosaccharideswith identical molecular weights, such as LNT andLNneoT, 2 *-FucLac and 3-FucLac or LNF-I from LNF-II and LNF-III. However, LNF-II and LNF-III coelute,or, in a few runs, only partially resolve.

Perbenzoylation has been used previously for the chro-matographic resolution of high-mannose oligosaccha-rides. Quantification of these oligosaccharides was basedon the peak area of absorbance at 229 nm. In theory, aseach hydroxyl group is replaced by a benzoyl group, theobserved total UV absorbance should reflect the additiveabsorbancies of all the benzoyl groups in the molecule.This was reported for the quantification of oligosaccha-

FIG. 1. Human milk oligosaccharide analysis. Defatted deprotei-nated samples are reduced, and neutral oligosaccharides are perben-zoylated, resolved by reversed-phase HPLC, and detected at 229 nm.Samples of 50 mg (10 mg of nonlactose oligosaccharides), equivalentto approximately 1 ml of human milk, give optimum chromatographicseparation and detection.

at 6 pmol of LNF-I, the S:N was 20:1. Thus, the methodis linear for representative oligosaccharides within therange usually found in milk samples (5 to 100 mg) andwell beyond (1 to 1000 mg lactose) through the entireextraction, perbenzoylation, clean-up, HPLC separa-tion, and detection steps. The limit of sensitivity is wellbelow that needed for routine analyses of human milksamples.

For routine analysis of milk samples, approximately50 mg of total oligosaccharides was injected into theHPLC. However, it should be noted that 80 to 90% ofthis was lactose. Thus, approximately 10 mg of nonlac-tose oligosaccharides gave an optimal response. Usingthis procedure, it was possible to clearly identify twelve FIG. 2. Peak area as a function of oligosaccharide concentration.major milk oligosaccharides ranging in size from tri- (A) Known amounts of lactose ranging from 1 mg to 1 mg were re-

duced with NaBH4 and perbenzoylated. HPLC was carried out on ato octasaccharides. Based on the published informationC-8 column using a gradient of acetonitrile and water with detectionavailable (7), these oligosaccharides would account forat 229 nm. The peak areas obtained by a Macintosh computer run-more than 70% of the total milk oligosaccharides by ning Dynamax peak integration software were plotted against lac-

weight. Figure 3A shows the profile obtained from a tose concentrations. For amounts of lactose less than 1 mg, the SDmixture of neutral milk oligosaccharide standards pro- error bars are within the demarcation of the symbol and not visible.

(B) A similar analysis of LNF-I was performed to ascertain thatcured commercially. The purity of each individual stan-perbenzoylation/HPLC analysis of larger milk oligosaccharides givesdard was confirmed by the HPLC analysis of their per-results comparable to those of lactose. Amounts of LNF-I from 5 tobenzoylated derivatives (Fig. 4) and was found to be 100 mg were treated as above and likewise produced a linear re-

over 90% for most of the standards, with the exception sponse. For amounts of LNF-I less than 100 mg, error bars are withinthe demarcation of the symbol.of LDFH-I, which contained 25% LDFT, and of LDFH-

AID AB 2250 / 6m3e$$$222 07-24-97 14:49:02 abal

CHROMATOGRAPHIC PROFILING OF MILK OLIGOSACCHARIDES 93

sults with the milk oligosaccharides differ in that as themolecular size of the oligosaccharides increased, the inte-grated peak areas did not correspond to those expectedfrom the number of benzoyl groups. This deviation fromlinearity increased with increasing size through DP-8,above which single peaks were not consistently detected.This disparity could have been due to incomplete benzo-ylation resulting from stearic hindrance in the large,highly branched molecules, with hindered hydroxylgroups being shielded from reaction by the bulky benzoylgroups already present. We have sometimes seen clus-ters of small peaks immediately preceding the majorpeak for DP-9 and DP-10 standards, indicating partialperbenzoylation; however, we did not see any such evi-dence for the DP-6, -7, and -8 compounds in milk samples(Fig. 3) or standards (Fig. 4). For perbenzoylated oligo-saccharides, the larger the oligosaccharide, the more hy-drophobic the molecule, raising the possibility that the

FIG. 3. (A) Oligosaccharide patterns of human milk. Known amountsof commercially obtained standard milk oligosaccharides were reducedwith NaBH4 and perbenzoylated. HPLC on a C-8 column used a gradi-ent of acetonitrile and water with detection at 229 nm. Molar absorptivi-ties were calculated by dividing the peak areas by concentrations ofthe standard oligosaccharide. (B and C) Aliquots of oligosaccharides(0.5 mg) from human milk samples were reduced with sodium borohy-dride and perbenzoylated. HPLC was carried out as in A. Oligosaccha-rides corresponding to peaks in the chromatograms were identified bycoelution with authentic standards and by electrospray mass spectrom-etry after O-debenzoylation with mild alkali.

FIG. 4. HPLC of individual perbenzoylated standard milk oligosac-rides where the absorbance was directly proportional to charides. Most standards displayed 90% purity, and there was nothe number of benzoyl groups independent of the size of evidence of the clusters of peaks immediately preceeding the major

peak that would indicate incomplete perbenzoylation.the high-mannose oligosaccharides tested (8). Our re-

AID AB 2250 / 6m3e$$$223 07-24-97 14:49:02 abal

CHATURVEDI ET AL.94

TABLE 2 resentative milk oligosaccharide. Total hexose was de-termined by the phenol–sulfuric acid procedure (9),Molar Absorptivity of Perbenzoylated Milk Oligosaccharideusing a calibration curve derived from known amountsStandardsa (peak area (mV-sec)/pmol)of a mixture of fucose, galactose, N-acetylglucosamine,

Oligosaccharide Mean SE and glucose in the proportions in which they occur inLNF I, i.e., 1:2:1:1. L-Fucose was determined by theLactose 2730 300procedure of Dische and Shettles (10) with a calibration2 *-FucLac 3520 130

3-FucLac 3530 50 curve derived from the same mixture. Amounts of LNF-LDFT 2940 150 I were then determined (i) by calculation based on hex-LNT and LNneoT 2840 100 ose content, (ii) by calculation based on fucose content,LNF-I, II, III 2440 110 and (iii) from HPLC peak areas and molar absorptivityLDFH-I, II 2150 120

of the perbenzoylation/HPLC method described herein.MFLNH-III 760 70DFLNHa 520 40 As can be seen from Table 4, there was generally good

agreement among the three procedures and the gravi-a n Å 8. metrically calibrated standards. The perbenzoylation/

HPLC procedure had values that most closely matchedthose resulting from the phenol/sulfuric acid total hex-ose analysis, while the results of the fucose analysispoor yields are a result of the derivatives remaining on

the C-18 cartridge during elution with acetonitrile. We most closely matched those of the standards. Thus, per-benzoylation/HPLC quantification of milk oligosaccha-tested this possibility by stripping the cartridge with a

more non-polar solvent, chloroform/methanol (2:1) after rides provides values comparable to other standard,well-established methods.the usual elution with acetonitrile. However, this step

did not result in any substantive additional recovery of Human milk oligosaccharide levels and their varia-larger oligosaccharides. Thus, the most likely possibility tion. The perbenzoylation/HPLC technique was ap-is that even when the molecule is fully benzoylated, the plied toward describing the oligosaccharides in theUV absorbance of some of the benzoyl groups is inter- milk of 50 donors. All milk was obtained 30 to 60 daysnally quenched by the stearically complex environment postpartum, to minimize any variation that might existwithin the molecule. We conclude, therefore, that the at different stages of lactation. The results of this sur-lower than expected peak areas result from two factors: vey are given in Table 5. The major oligosaccharidesStearic hindrance to complete benzoylation above DP- are the two monofucosylated lactoses 2 *- and 3-FucLac,8 and internal quenching of UV absorbance for DP-8 the three monofucosylated pentasaccharides LNF-I, II,(DFLNHa) and smaller oligosaccharides. and III, and the octasaccharide DFLNHa. Combined,

The loss of linearity with increasing size could be they account for approximately three-quarters of thereadily compensated for by deriving molar absorptivi- oligosaccharides identified in our profiles. The valuesties for each of the identified oligosaccharides. Authen- of 2 *-FucLac range from 2.3 to 0.2 mg/ml (mean, 1.2tic standards were obtained commercially, perbenzoyl- mg/ml). The total calculated yields of oligosaccharideated, and quantified by this method. As can be seen in peaks obtained by this procedure corresponded to 80 toTable 2, the molar absorptivities are similar for iso- 90% of the starting weight of the total oligosaccharidemeric oligosaccharides, showing that the response ismore a function of molecular size than structural fea-tures. Most importantly, the molar absorptivities are

TABLE 3reproducible in different runs, and were not affectedIntraassay Variation in Human Milk Oligosaccharideby the addition of increasing amounts of lactose (25 to

Profile of a Single Milk Samplea (nmol/ml)400 mg) to a set of neutral milk oligosaccharide stan-dards (data not shown). Using the molar absorptivities

Oligosaccharide Mean SE CVb (%)obtained from authentic standards, the concentrationof each of the oligosaccharides identified in this study 2 *-FucLac 1740 80 13

3-FucLac 1700 69 11of human milk samples was calculated. The intraassayLDFT 65 5 20variability was determined by analyzing eight identicalLNT 197 9 13milk samples from a single donor. As seen in Table 3, LNneoT 158 9 15

the coefficient of variation of the oligosaccharide peaks LNF-I 976 42 12is generally in the 10 to 15% range. LNF-II / III 718 32 13

LDFH-I 702 37 15In order to confirm that oligosaccharides can be reli-LDFH-II 221 9 12ably quantified from HPLC areas and molar absorptivi-MFLNH-III 425 13 8ties, i.e., to ascertain that the derived values were truly

representative of the actual amounts present, a three- a n Å 8.b CV Å SD/mean 1 100.way comparison was performed, using LNF-I as a rep-

AID AB 2250 / 6m3e$$$223 07-24-97 14:49:02 abal

CHROMATOGRAPHIC PROFILING OF MILK OLIGOSACCHARIDES 95

TABLE 4

Perbenzoylation/HPLC Quantitation of LNF-I Standards Compared with Classic Colorimetry

Total sugarsStandard [phenol-sulfuric acid] Total fucose

[gravimetric] (Dubois et al. (9); (Dische & Shettles (10); Perbenzoylation/HPLC(mg LNF-I) mean % measured, n Å 2) mean % measured, n Å 3) (mean % measured, n Å 2)

5 87 103 8310 79 104 8420 86 100 8730 85 103 9850 81 104 78

fraction. Thus, the concentrations of individual oligo- 3. Fucosylated oligosaccharides are retained longerthan their nonfucosylated counterparts. For example,saccharides could be calculated without the use of in-the trisaccharides 2 *-FucLac and 3-FucLac show aternal standards.longer retention time than the tetrasaccharides LNTLarge variations in oligosaccharide concentrationsand LNneoT. It suggests that the increased hydropho-were observed among samples. One well-establishedbicity due to the methyl group of fucose more thansource of such variation in milk oligosaccharides re-compensates for the loss of hydrophobicity due to therelates to the secretor status and Lewis blood group typebeing one less benzoyl group. Also, the more polar na-of the mother, which interact to cause variation in theture of the GlcNAc residue in tetrasaccharides reducedmaternal expression of the a1,2 fucosyltransferasetheir retentivity under reversed-phase conditions.gene, and are known to result in significantly less pro-

4. Amongst perbenzoylated fucosyloligosaccharides,duction of fucosyloligosaccharides containing 2-linkedthe 2-linked structures (2 *-FucLac, LNF I) chromato-fucose in the milk of recessive mothers (11–14). In agraph earlier than their 3- or 4-linked counterparts,population derived from Europeans, typically 30% ofe.g., 3-FucLac or LNF-II and LNF-III, on C-8 columns.mothers express recessive patterns of milk oligosaccha-This contrasts with the reversed-phase HPLC of nativerides. Two oligosaccharide profiles shown in Figs. 3Boligosaccharides on a C-18 column using water as theand 3C are representative of the dichotomy betweeneluent, where the 2-linked or the straight chain struc-typical and atypical oligosaccharide profiles in our pop-tures are retained longer on the column (15–17).ulation, whose secretor status is undefined. Note the

5. The effect of fucose and branching is well illus-lack of oligosaccharides with 2-linked structures in thetrated by lactodifucotetraose (see structure, Table 1).atypical pattern shown in Fig. 3C.The combination of two fucose residues and a branchedTable 6 exhibits the mean values of oligosaccharidesstructure makes this compound interact more stronglyin our population alongside other published values ob-with the nonpolar C-8 stationary phase, resulting in atained by alternative methods of analysis. There is gen-retention time which is longer than that for the fucosy-erally good agreement between the mean concentrationslated pentasaccharides.of individual oligosaccharides among these studies, indi-

Conclusions. The reversed phase HPLC of perbenzo-cating that they are all accurate depictions of milk oligo-ylated oligosaccharides of human milk is a straightfor-saccharide profiles in the human population.

Chromatographic patterns. The chromatographic TABLE 5behavior of perbenzoylated milk oligosaccharides on re-

Oligosaccharides of Human Milka (mg/ml)versed-phase HPLC follows these patterns:Oligosaccharide Mean SE

1. Oligosaccharides generally elute in the order ofincreasing molecular weight. 2 *-FucLac 1.21 0.08

3-FucLac 0.36 0.042. The retention times of oligosaccharides of identi-LDFT 0.30 0.06cal molecular weight are influenced by the linkage ofLNT 0.40 0.05sugar at the nonreducing terminus. For example, LNneoT 0.18 0.02

LNneoT [containing Galb(1 r 4)GlcNAc] elutes slightly LNF-I 0.94 0.12earlier than LNT [containing Galb(1 r 3)GlcNAc]. Ste- LNF-II / III 0.64 0.08

LDFH-I 0.55 0.11reochemistry of the oligosaccharide plays an importantLDFH-II 0.11 0.01role in differential chromatographic behavior. This is inMFLNH-III 0.34 0.04agreement with earlier findings (8) of chromatographic DFLNHa 0.93 0.16

separation of anomeric forms of isomaltose, isomalto-a n Å 50.triose, isomaltotetraose, etc.

AID AB 2250 / 6m3e$$$223 07-24-97 14:49:02 abal

CHATURVEDI ET AL.96

TABLE 6

Comparison among Reports of Oligosaccharide Levels in Human Milk (mg/ml)

TLC, HPAEC HPAEC(Kunz et al. (7)) (Thurl et al. (5)) Perbenzoylation/HPLC

Number of milk samples: Unspecified 1 50Stage of lactation: Unspecified Unspecified 30–60 days

LNneoT 0.1 0.2LNT 0.5–1.5 0.9 0.42 *-FucLac 1.8 1.23-FucLac 0.5 0.4LNF-I 1.0–1.5 0.7 0.9LNF-II / III 0.5–1.0 0.5 0.6LDFT 0.2 0.3LDFH-I 0.6 0.6LDFH-II 0.2 0.1MFLNH-III 0.3 0.3DFLNHa 0.2 1.3Total oligosaccharides 3–6 5.9 6.3

ward, reliable, and sensitive method to obtain qualita- tigate neutral and acidic oligosaccharides of human milk(3). This technique is undoubtedly very useful, especiallytive as well as quantitative information about milk

oligosaccharides. Over 250 milk samples have been run for larger oligosaccharides. However, it is an emergingtechnique, and access to the sophisticated and expensivesuccessfully to date without failure and with consistent

results, and the sensitivity is at least three orders of instrumentation may be limited. Furthermore, it be-comes almost essential to eliminate lactose prior tomagnitude greater than any other published chromato-

graphic method. Although the method is suitable only structural analysis of oligosaccharides. A recently re-ported method of Thurl et al. (5) using high-pH anion-for oligosaccharides through DP-8, the relative amounts

of these specific smaller oligosaccharides could predict exchange chromatography also requires lactose to beremoved prior to quantification of milk oligosaccharides.the occurrence of corresponding higher structural ana-

logs because their synthesis most likely involves the In contrast, the perbenzoylation method describedherein does not require a chromatographic step for thesame glycosyltransferases.

Human milk oligosaccharides play a protective role specific removal of lactose. Other advantages of the per-benzoylation method lie in its sensitivity, reproducibil-against various bacterial and viral pathogens (18–20).

The relationship between variation in oligosaccharide ity, and the simplicity of detection. Moreover, it is possi-ble to hydrolyze the benzoyl groups from derivatizedprofiles in individual milk samples and protection in

the breast-fed infant against oligosaccharide-sensitive oligosaccharides to get back their debenzoylated formand use them for further studies. These considerationspathogens is of interest, as such a relationship would

allow predictions of differential susceptibility to vari- make the reversed-phase HPLC of perbenzoylated milkoligosaccharides a method of choice for the study of oligo-ous pathogens among breast-feeding infants. Earlier

work carried out in our laboratory (6) showed varia- saccharide profiles in milk samples.tion in the activities of fucosyltransferase and fucosi-dase in milk samples from different donors and at

ACKNOWLEDGMENTSdifferent lactational stages. Studies by Miller et al.(4) on the monosaccharides derived mainly from milk Dedicated to Roger W. Jeanloz on the occasion of his 80th birthday.oligosaccharides also indicate large intraindividual The authors thank Gherman Ya. Wiederschain and Stephen C. Lui

for assistance with this project and Roger W. Jeanloz for his input.and temporal changes in the oligosaccharide composi-We are also grateful to Catherine E. Costello, Vernon Reinhold, Songtion of human milk. These issues of individual varia-Ye, and Bruce Reinhold of the Boston University Mass Spectrometrytion and protection against pathogens can now be ad- Resource for their mass spectrometric analysis and interpretation of

dressed directly by the method described herein, as it the oligosaccharide peaks.provides both qualitative and quantitative informa-tion on intact milk oligosaccharides. The considerable

REFERENCESvariation of milk oligosaccharide profiles over thecourse of lactation will be the subject of a subsequent 1. Kobata, A., Ginsburg, V., and Tsuda, M. (1969) Arch. Biochem.report. Biop. 130, 509–513.

Recently, matrix-assisted laser desorption/ionization 2. Newburg, D. S. (1996) J. Mammary Gland Biol. Neoplasia 1,271–283.time-of-flight mass spectrometry has been used to inves-

AID AB 2250 / 6m3e$$$223 07-24-97 14:49:02 abal

CHROMATOGRAPHIC PROFILING OF MILK OLIGOSACCHARIDES 97

3. Stahl, B., Thurl, S., Zeng, J., Karas, M., Hillenkamp, F., Steup, 12. Viverge, D., Grimmonprez, L., Cassanas, G., Bardet, L., and Sol-ere, M. (1986) Ann. Nutr. Metab. 30, 196–209.M., and Sawatzki, G. (1994) Anal. Biochem. 223, 218–226.

4. Miller, J. B., Bull, S., Miller, J., and McVeagh, P. (1994) J. Pedi- 13. Viverge, D., Grimmonprez, L., Cassanas, G., Bardet, L., and Sol-ere, M. (1990) J. Pediatr. Gastroenterol. Nutr. 11, 365–370.atr. Gastroenterol. Nutr. 19, 371–376.

5. Thurl, S., Muller-Werner, B., and Sawatzki, G. (1996) Anal. Bio- 14. Viverge, D., Grimmonprez, L., Cassanas, G., Bardet, L., and Sol-ere, M. (1990) J. Pediatr. Gastroenterol. Nutr. 11, 361–364.chem. 235, 202–206.

6. Wiederschain, G. Y., and Newburg, D. S. (1995) J. Nutr. Bio- 15. Cheetham, N. W. H., and Dube, V. E. (1983) J. Chromatogr. 262,426–430.chem. 6, 582–587.

7. Kunz, C., and Rudloff, S. (1993) Acta Paediatr. 82, 903–912. 16. Dua, V. K., and Bush, C. A. (1983) Anal. Biochem. 133, 1–8.17. Dua, V. K., Dube, V. E., and Bush, C. A. (1984) Biochim. Biophys.8. Daniel, P. F. (1987) in Methods in Enzymology (Ginsburg, V.,

Ed.), Vol. 138, pp. 94–116, Academic Press, San Diego. Acta 802, 29–40.18. Newburg, D. S., Pickering, L. K., McCluer, R. H., and Cleary,9. Dubois, M., Gilles, K. A., Hamilton, J. K., Rebers, P. A., and

Smith, F. (1956) Anal. Chem. 28, 350–356. T. G. (1990) J. Infect. Dis. 162, 1075–1080.19. Andersson, B., Porras, O., Hanson, L. A., Lagergard, T., and10. Dische, Z., and Shettles, L. B. (1948) J. Biol. Chem. 175, 595–

603. Svanborg-Eden, C. (1986) J. Infect. Dis. 153, 232–237.20. Holmgren, J., Svennerholm, A.-M., and Lindblad, M. (1983) In-11. Viverge, D., Grimmonprez, L., Cassanas, G., Bardet, L., Bonnet,

H., and Solere, M. (1985) Ann. Nutr. Metab. 29, 1–11. fect. Immun. 39, 147–154.

AID AB 2250 / 6m3e$$$224 07-24-97 14:49:02 abal