Remodeling of mouse milk glycoconjugates by transgenic expression of a human glycosyltransferase.

The mammary gland is a unique biosynthetic tissue that produces a variety of species-specific glycoconjugates, but the factors regulating the production of specific glycoconjugates are not well understood. To explore the underlying regulation, a fusion gene containing a cDNA encoding the human alpha 1,2-fucosyltransferase (alpha 1,2FT), which generates the H-blood group antigen, flanked by the murine whey acidic protein promoter and a polyadenylation signal, was introduced into mice. Milk samples from transgenic animals contained soluble forms of the alpha 1,2FT, as revealed by Western blots of milk samples using an anti-alpha 1,2FT antiserum and by the demonstration of alpha 1,2FT enzyme activity. Milk from transgenic animals also contained large quantities of 2'-fucosyllactose (Fuc alpha 1-2Gal beta 1-4Glc) and modified glycoproteins containing the H-antigen, whereas milk from control animals lacked these glycoconjugates. Expression levels of 2'-fucosyllactose were high in most animals and represented 1/3 to nearly 1/2 of the total milk oligosaccharides. These results demonstrate that heterologous transgenic expression of a glycosyltransferase can result in the expression of both the transgene and its secondary gene products and that the structures of milk oligosaccharides can be remodeled depending on expression of the appropriate enzyme. Furthermore, these results suggest that the lactating mammary gland may be a unique biosynthetic reactor for the production of biologically active oligosaccharides and glycoconjugates.

ficult. One approach that has proven somewhat successful in vitro is to attempt to alter oligosaccharide biosynthesis by genetic manipulation of glycosyltransferases (Lee et al., 1989;Smith et al., 1990;Lowe et al., 1990). This approach has so far been limited to cultured and immortalized cell lines and has been useful in producing neoglycoconjugates for biological studies .
One major site for glycoconjugate biosynthesis in animals, and one that has been of enormous historical importance in the field of glycobiology, is the lactating mammary gland. Milk contains, in addition to lactose, numerous glycoproteins (Patton et al., 1990;Fiat and Jollès, 1989) and a variety of free oligosaccharides (Kobata et al., 1972). The structure and composition of the free oligosaccharides in milk differ between mammals, but the human lactating mammary gland synthesizes the most complex mixture of free, reducing oligosaccharides, many of which contain fucose and other determinants for human blood groups, including the ABO and Lewis system. The structures of some of these oligosaccharides in humans are determined genetically by the expression of glycosyltransferases during lactation and the availability of sugar nucleotide substrates (Hill and Brew, 1975;Ginsburg, 1972;Smith et al. 1987), but the precise mechanisms regulating differential expression of individual milk oligosaccharides are not known.
Recent advances in transgenic technology have resulted in successful expression of transgene-encoded proteins in animal milk (Wilde et al., 1992;Paleyanda et al., 1991;Archer et al., 1994). Using this methodology, we have begun to explore factors regulating glycoconjugate biosynthesis. Our analyses of normal mouse milk demonstrated that it is deficient in fucosecontaining oligosaccharides, and in particular it lacks oligosaccharides containing the H-blood group antigen. Based on this information, we chose to use a cDNA encoding a human ␣1,2fucosyltransferase (␣1,2FT) 1 responsible for synthesis of the blood group H-antigen Fuc␣1-2Gal␤1-3(4)-R (Rajan et al., 1989;Larsen et al., 1990). Our studies demonstrate that the human ␣1,2FT can be successfully introduced into transgenic animals causing the lactating mammary gland to synthesize relatively large quantities of oligosaccharides containing the H-antigen and at least one major glycoprotein containing this determinant. These studies provide new insights into the factors regulating biosynthesis of milk oligosaccharides within the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed: Ross Laboratories, 3300 Stelzer Rd., Columbus, lactating mammary gland and a new approach for studying factors controlling glycoconjugate biosynthesis.

EXPERIMENTAL PROCEDURES
Preparation of Vectors Containing the Human ␣1,2FT-The human ␣1,2FT was cloned by polymerase chain reaction from an epidermal carcinoma cell line (A431) cDNA library using primers derived from the previously published ␣1,2FT gene sequence . Direct sequencing of the ␣1,2FT cDNA confirmed its identity with the known sequence. To demonstrate that the cDNA encoded a functional ␣1,2FT enzyme, the ␣1,2FT cDNA was inserted into an expression plasmid (pMet-FucT-bGH) and introduced into mouse L cells, which do not contain either ␣1,2FT enzyme activity or H-blood group antigens. This plasmid utilizes the mouse metallothionein-I promoter (Bayne et al., 1987) to direct expression and contains the bovine growth hormone polyadenylation signal (bGH poly(A)) for proper mRNA processing. Recombinant enzyme activity was demonstrated in the mouse L cells by indirect immunofluorescent staining of positive transfectants with Ulex europaeus agglutinin I (UEA I) (Sigma), which reacts with H-blood group antigen, and a monoclonal antibody against H-blood group antigen (Dako Corp.). To generate antibody to the ␣1,2FT, the cDNA encoding the ␣1,2FT was subcloned into an inducible expression vector (pQE11) (Qiagen, Inc.) in frame with a 6 ϫ histidine tag. This 6 ϫ histidine-tagged protein was purified by nickel affinity chromatography and injected in rabbits to generate polyclonal antibodies to the ␣1,2FT.
For transgenic studies, the cDNA fragment encoding the ␣1,2FT was inserted into a second expression plasmid, pWAP-␣1,2FT (Fig. 1). This plasmid contains bGH poly(A) and utilizes the murine whey acidic protein (WAP) promoter (Pittius et al., 1988) to direct gene expression primarily to lactating mammary gland tissue. An EcoRI-BamHI DNA fragment containing the WAP-␣1,2FT-bGH poly(A) sequences was isolated from this plasmid and injected into fertilized mouse (B6/SJL) eggs as described previously (McGrane et al., 1988). The WAP promoter was generously provided by Dr. Lothar Hennighausen (National Institutes of Health).
Preparation of Milk Samples-Individual milk samples were obtained from transgenic and non-transgenic mothers after nursing the pups for 5-12 days. Samples were collected using a mammary suction apparatus consisting of a 23-gauge needle and a vacuum line connected to a trap test tube. Samples were frozen on dry ice and stored at Ϫ70°C. Oligosaccharide extracts were prepared using a method modified from Kobata et al. (1972). Typically, 50 -100-l milk aliquots were centrifuged for 15 min at 11,000 ϫ g. The resulting lipid pads were removed from the top, and the liquid layer was transferred to a clean vial. A 2-fold volume of cold ethanol was added with mixing, and the samples were then centrifuged as before. The clear supernatants were transferred to clean vials and dried for further analysis. Precipitates, containing the proteins, were dried and stored at Ϫ70°C until used for further analysis.
Isolation and Separation of Oligosaccharides-Oligosaccharides were separated, detected, and quantified by high performance anion exchange chromatography (HPAEC) and fluorophore-assisted carbohydrate electrophoresis (FACE). HPAEC was performed by a method adapted from Reddy and Bush (1991). Oligosaccharide standards for HPAEC and FACE were obtained from V-Labs Inc. (Covington, LA).
Individually dried oligosaccharide extracts from mouse milk, prepared as above, were resuspended in water to the original volume of their respective milk samples. Aliquots (20 l) were injected onto a Dionex CarboPac-PA1 analytical column. Analytes were detected using a pulsed electrochemical detector equipped with a gold working electrode and a pH reference electrode. Elution was started with a solution of 100 mM NaOH (RICCA Chemicals, Arlington, TX) held steady for 12 min and changed to 7% of 600 mM sodium acetate in 100 mM NaOH. After 8 min, this percentage was increased to 10% and held constant for 7 min. Flow rate was maintained at 1 ml/min. The column was washed in a mixture of 50% 100 mM NaOH, 50% 600 mM sodium acetate in 100 mM NaOH and equilibrated for 15 min in starting solution before reuse. After detection, the elution stream was desalted using a Dionex AMMS-II anion micromembrane suppresser fitted with an AutoRegen apparatus equipped with an anion exchange cartridge and 0.15% H 2 SO 4 regenerant solution. Fractions (0.5 ml) were collected and immediately frozen for lyophilization. Fractions obtained by HPAEC (or glycosidase digests from these fractions) were submitted to FACE analysis (Jackson, 1990;Stack and Sullivan, 1992). Electrophoresis equipment, imager, labeling, and reducing reagents for fluorophore labeling were from Glyko Inc. (Novato, CA). Manufacturer's procedures were followed.
Exoglycosidase Treatments-Fucosidases from Corynebacterium sp. (␣1-2 specific) and Streptomyces (␣1-3/4 specific) were from Takara Panvera (Madison, WI), and Escherichia coli galactosidase was from Boehringer Mannheim. All other chemicals used were from the highest purity available and, unless otherwise indicated, were from Sigma.
Immunoblotting of Milk Proteins-The milk protein pellets obtained by ethanol precipitation as described above were tested for the presence of the ␣1,2FT polypeptide by immunoblot using antisera raised against the recombinant enzyme. Pellets were resuspended in a volume of SDS equal to that of the original volume of the milk sample. Aliquots (5 l) of resuspended pellets were electrophoresed on 12.5% polyacrylamide gels (Laemmli, 1970). Resolved proteins were transferred to nitrocellulose membranes for 1 h at 100 volts. The resulting membranes were blocked by incubation in 50 mM TBS, pH 7.5, containing 0.5 M NaCl and 2% gelatin. Membranes were washed three times in TBS containing 0.05% Tween 20 and were incubated for 18 h in 1% gelatin/TBS containing 1:500 dilution of rabbit polyclonal anti-␣1,2FT. Membranes were then incubated with 1% gelatin/TBS containing goat anti-rabbit IgG previously conjugated to horseradish peroxidase and washed with TBS-Tween. Proteins were visualized by incubating in 50 ml of TBS containing 0.018% hydrogen peroxide and 10 ml of methanol containing 30 mg of 4-chloro-napthol.
Lectin Blotting of Milk Proteins-Blotted proteins on nitrocellulose membranes were probed for the presence of ␣1,2-linked fucose by incubating for 18 h in a 2% gelatin/TBS containing a 1:500 dilution of the peroxidase labeled UEA-I (Sigma). The resulting membranes were then washed, and proteins were visualized by incubating in a mixture of 50 ml of TBS containing 0.018% hydrogen peroxide and 10 ml of methanol containing 30 mg of 4-chloronapthol (Bio-Rad).

Analysis of Oligosaccharides in Milk from Control Mice-To
determine a suitable glycosyltransferase for mammary glanddirected transgenic expression in mice, we analyzed mouse milk oligosaccharides using HPAEC. Mouse milk contains lactose, sialyllactoses, and other oligosaccharides (Fig. 2, panel A). One of the minor oligosaccharides of murine milk coeluted with authentic 3-fucosyllactose (3-FL, Gal␤1-4(Fuc␣1-3)Glc). Its identity was further confirmed by fucosidase digestion and FACE analysis (data not shown). This demonstrates that mouse milk contains detectable but low levels of 3-FL. 2Ј-Fucosyllactose (Fuc␣1-2Gal␤1-4Glc), which is an isomer of 3-FL, was not detected in the milk of control animals. In a separate experiment, a milk sample was exhaustively treated with a mixture of fucosidases, and the amount of released fucose was determined by HPAEC. Mouse milk was found to contain approximately 1 nmol of fucose per 100 l. Although the amounts of total fucose and 3-FL in mouse milk are minuscule, their presence indicates that lactating mouse mammary glands most likely contain GDP-Fuc, which is the donor sugar nucleotide for fucosyltransferases.
Expression of the Human ␣1,2FT in Transgenic Mice-The absence of 2Ј-FL in mouse milk suggests that the lactating mammary gland of the mouse lacks ␣1,2-fucosyltransferase activity. A fusion gene containing a cDNA encoding the human ␣1,2FT was prepared and injected into mouse embryos as described under "Experimental Procedures." A total of 85 pups were obtained from these injections, and 9 of these animals (8 male and 1 female) were found to contain at least one copy of the transgene by quantitative slot blot hybridization (McGrane et al., 1988). Six of the transgenic animals (all male) were mated to normal mice, resulting in a total of 98 F 1 progeny. 38 of these animals (19 males/19 females) contained the ␣1,2FT gene. 15 of the females were bred with normal mice. Subsequent generations (F 2 and F 3 ) were obtained by interbreeding transgenic animals. Milk samples from transgenic and control non-transgenic animals were either directly tested for enzyme activity or, alternatively, treated with ethanol to obtain oligosaccharide extracts and protein pellets described under "Experimental Procedures." Milk samples from transgenic animals contained detectable ␣1,2FT activity as determined by a method modified from Rajan et al. (1989), while activity was absent in samples from control animals (data not shown). These results demonstrate that active ␣1,2FT enzyme was expressed in milk from transgenic animals.
The presence of ␣1,2FT polypeptide was determined by immunoblot analysis of milk protein using a polyclonal rabbit antibody against ␣1,2FT, as shown in Fig. 3. No immunoreactive protein was present in milk from control animals, whereas milk from transgenic animals contained four immunoreactive species of apparent molecular mass of 46, 32, 30, and 25 kDa (Fig. 3). The ␣1,2FT cDNA encodes the full-length type-2 transmembrane form of the enzyme that has a predicted size of approximately 46 kDa when fully glycosylated. Whether the 46-kDa form of the ␣1,2FT observed in transgenic milk represents a full-length form of the enzyme or a partly degraded form with altered glycosylation is not yet known. Nevertheless, the presence of discrete lower molecular weight forms of the enzyme in milk indicates discrete proteolytic cleavages of the polypeptide resulting in the species observed. These lower molecular weight forms were not found in cell extracts of ␣1,2FT transfected mouse L-cells. The potential proteolytic cleavage sites of the recombinant enzyme are under investigation.
Presence of Fuc␣1-2Gal␤1-R Determinants in Milk Oligosaccharides from Transgenic Animals-Oligosaccharide extracts of milk samples from transgenic mice were analyzed for the presence of fucosylated oligosaccharides expected to be synthesized by the expressed ␣1,2FT. Milk oligosaccharides from two transgenic animals were analyzed by Dionex-HPAEC and compared to those in milk from a control mouse (Fig. 2, panels B and C). A major neooligosaccharide found in the milk of the FIG. 2. Chromatographic profiles of mouse milk oligosaccharides. Oligosaccharide extracts from a control non-transgenic mouse and two different transgenic mice, both of which were descendants of Founder 28, were subjected to Dionex high performance liquid chromatography. A, milk from control mice did not contain any DFL (a2 region) or 2Ј-FL (a3). A trace of 3-FL (a1) was detected and analyzed as described in the text. B, a second generation transgenic mouse (code 28-28) synthesized a large amount of 2Ј-FL (b3). Traces of 3-FL (b1) and DFL (b2) were detected. C, a larger amount of 2Ј-FL was present in milk from a homozygous third generation transgenic animal (code 28-28-37). The elution positions of monosaccharides and standard oligosaccharides are indicated. FIG. 3. ␣1,2FT was present in milk of transgenic animals. Protein pellets from milk samples were separated by SDS electrophoresis, blotted, and assayed for binding to anti-␣1,2FT antiserum as described under "Experimental Procedures." Milk proteins from control animals were not specifically recognized by the antiserum (lane 1), while some proteins from transgenic animals (lanes 2 and 3) were recognized. One of the detected bands (identified by an arrow) was in the 46-kDa region, which corresponds to the expected molecular weight of the full-length ␣1,2FT. The migration positions of molecular weight markers are indicated.
transgenic animals eluted on Dionex-HPAEC identically to authentic 2Ј-FL. Analysis of milk from other donors revealed that this neooligosaccharide was present in all milk samples obtained from transgenic animals of all generations with the exception of one F 1 donor.
To confirm that the neooligosaccharide in the milk of transgenic animals was 2Ј-FL, the purified oligosaccharide collected after Dionex-HPAEC was subjected to FACE analysis and specific enzyme degradation. An electrophoretogram of the neooligosaccharide before and after treatment with ␣1,2-specific fucosidase from Corynebacterium sp. is shown in Fig. 4. The fluorophore-labeled neooligosaccharide (lane 4) comigrated with authentic 2Ј-FL (lane 2), and, like the standard 2Ј-FL, it was susceptible to the action of the ␣1,2-specific fucosidase (lanes 3 and 5). The neooligosaccharide was also exhaustively treated with a combination of the ␣1,2-specific fucosidase and E. coli ␤-galactosidase. This combined treatment resulted in the release of equimolar amounts of fucose, galactose, and glucose, as determined by FACE analysis (data not shown). These results demonstrate that the neooligosaccharide present in the milk of transgenic animals expressing the human ␣1,2FT is 2Ј-FL.
A minor oligosaccharide that coelutes with the human milk tetrasaccharide difucosyllactose (DFL, Fuc␣1-2Gal␤1-4(Fuc␣1-3)Glc) was also detected in the chromatographic profiles from transgenic animals (Fig. 2, panels B and C). Fractions corresponding to the elution time of this oligosaccharide were pooled and analyzed by FACE. The resulting band obtained from transgenic animal milk comigrated with authentic DFL (data not shown). Results were consistent with the presence of this oligosaccharide only in milk from transgenic animals. It is expected that this product is generated by the sequential action of the ␣1,2FT and the endogenous ␣1,3FT.
Presence of Glycoproteins Containing the H-blood Group Determinant in Milk of Transgenic Animals-The milk proteins of transgenic and control animals were separated by SDS-poly-acrylamide gel electrophoresis and transferred to nitrocellulose membranes by electroblotting. These blots were then assayed for the presence of Fuc␣1-2Gal␤1-R determinants using the plant lectin UEA-I, which displays high affinity for H-blood group determinants (Allen et al., 1977;Hindsgaul et al., 1982). The milk of control animals lacked glycoproteins reactive with UEA-I, whereas milk of transgenic animals contained a major reactive glycoprotein of apparent molecular mass of 35-40 kDa in addition to a minor reactive band of 29 kDa (Fig. 5). This pattern is different from that observed for the recombinant ␣1,2FT (Fig. 3). This major fucosylated glycoprotein may be casein, but this is under investigation. These results demonstrate at least one major glycoprotein is an acceptor for the transgenically expressed human ␣1,2FT and becomes highly modified by the enzyme, whereas other glycoproteins in the milk appear to be less efficiently modified. Table I is a summary of results for control and transgenic animals generated by different founders and belonging to different generations. Transgenic mice obtained from founders 28 and 34 (Table I) produced typical amounts of 2Ј-FL (average 2.2 g/liter). In one case, the amount of 2Ј-FL (Ho-29 -129) was almost one-half of the total milk oligosaccharides, demonstrating that it was present at a level approximately equal to that of lactose. The transgene has been stable through the germ line for three generations as reflected by the ability to synthesize 2Ј-FL. Interestingly, the ability to synthesize ␣1,2-fucosylated glycoconjugates seems to be independent of the number of transgene copies as determined by tail biopsy analyses (Table I). DISCUSSION Results presented in this report demonstrate that it is possible to dramatically alter the structure of the milk oligosaccharides and glycoproteins by the transgenic introduction of a heterologous glycosyltransferase under the control of a lacto- FIG. 4. The major neooligosaccharide found in milk of transgenic mice comigrated with authentic 2-FL in FACE analysis and was susceptible to a specific ␣1,2-fucosidase. Fractions corresponding to the major neooligosaccharide (region b3, Fig. 2, panel B) were pooled. An aliquot was directly labeled with 8-aminonapthalene-2,3,6-trisulfonic acid and subjected to FACE analysis (lane 4). A second aliquot was treated with 20 milliunits of Corynebacterium sp. ␣1-2 fucosidase prior to labeling (lane 5). Authentic 2Ј-FL was also labeled (lane 2) and treated with the enzyme (lane 3). Lane 1 contains labeled oligosaccharide standards.

Overall Expression of 2Ј-FL in Milk of Transgenic Animals-
FIG. 5. Glycoproteins containing the novel Fuc␣1-2 linkages were detected in milk from transgenic animals. Protein pellets from mouse milk samples were resuspended in a volume of SDS-polyacrylamide gel electrophoresis sample buffer equal to that of the original volume of milk. Aliquots of these extracts (5 l) were resolved by electrophoresis and blotted in nitrocellulose membranes. The membranes were then incubated with a preparation of labeled UEA I, as described under "Experimental Procedures." Samples obtained from transgenic animals (lanes 2-5) were specifically recognized by UEA I. Fucose ␣1,2 linkages were not detected in samples from control animals (lanes 1 and 6). The migration positions of molecular weight markers are indicated.
genically induced expression of the enzyme. We chose the human ␣1,2FT as a model enzyme to explore its expression within the mammary gland and the consequences of its expression on biosynthesis of milk glycoconjugates. We found that 2Ј-FL, which is lacking in normal mouse milk, was a major oligosaccharide in the milk of transgenic animals and was present at levels rivaling that of lactose. In addition, active human ␣1,2FT and an oligosaccharide that coelutes with difucosyllactose were found in the milk, and at least one milk glycoprotein was a major substrate for the transgenic ␣1,2FT.
Biosynthesis of milk oligosaccharides from lactose would be expected to be directly related to the expression of particular glycosyltransferases, but other factors, such as sugar nucleotide availability, lactose availability, enzyme localization, enzyme secretion, etc., could profoundly alter biosynthesis. We reasoned that the experimental introduction of the transgenic ␣1,2FT would test many of these factors. For example, we considered the possibility that the low levels of 3-FL, the absence of 2Ј-FL, and the low amount of fucose in mouse milk might be due to a low expression of GDP-Fuc. The high level expression of 2Ј-FL we observed, however, indicates that the Golgi apparatus of the lactating mammary gland must be sufficiently adaptable in regard to GDP-Fuc uptake to accommodate the need of the transgenic ␣1,2FT. We also considered the possibility that the transgenic ␣1,2FT might be proteolytically processed within the lactating mammary gland and be either inactivated or quantitatively secreted without being functionally present in the Golgi apparatus. It is conceivable that some other glycosyltransferase genes expressed as transgenes within the lactating mammary gland will not be functional and be unable to cause the efficient remodeling of milk glycoconjugates. Further studies with other genes are required to address the generality of this phenomenon.
All samples from homozygous animals analyzed so far contained more 2Ј-FL than those of hemizygous descendants of the same founder. There were, however, variations of 2Ј-FL content among milk samples from hemizygous animals descending from the same founder. We have found that in humans, milk oligosaccharide content varies between samples from the same donor (data not shown). In mice, lactation stage, number of pups, age, and nutritional status are some of the factors that may contribute to the observed differences.
The functions of milk oligosaccharides are not known, but free oligosaccharides are known to influence the adhesion of pathogenic bacteria to human tissue in vivo (Andersson et al., 1986), suggesting that milk oligosaccharides may have a protective function in the infant. The ability to remodel milk oligosaccharides by the heterologous expression of glycosyltransferase genes should facilitate future experiments aimed at defining the biological functions of the lactose-derived oligosaccharides.
In addition, the transgenic expression of glycosyltransferases in lactating mammary glands presents a unique opportunity to develop technology that may result in the synthesis of large amounts of neooligosaccharides and substantially modified glycoproteins with therapeutic or industrial value. The findings described in this report may constitute a new method for the production of custom designed homologous or heterologous glycoconjugates. a Samples from control (CD1, Balb, B6SJL) animals did not contain 2Ј-fucosyllactose, while samples from transgenic animals had varying amounts of the trisaccharide.
b He and Ho represent animals hemizygous or homozygous, respectively, for the 1,2FT. c The first number in the identifying codes for transgenic animals indicates founder (F 0 ); second and third numbers identify specific transgenic mice in the progeny.