Butyrophilin Is Expressed in Mammary Epithelial Cells from a Single-sized Messenger RNA as a Type I Membrane Glycoprotein*

We investigated the expression of butyrophilin in eu-karyotic cells with a view to determining the number of mRNA species, the incorporation of the peptide chain into microsomes, and the topology of the processed protein in biological membranes. Butyrophilin is synthesized from a single sized mRNA in both bovine and murine lactating mammary tissue and associates with microsomal membranes with a type I topology (N exo z C cyto ) via a single hydrophobic anchor in the mid- dle of the sequence. Several isoelectric variants of the protein were detected in cellular membranes from lactating bovine mammary tissue and in the milk-fat-glob-ule membrane. We found no evidence for soluble forms of butyrophilin in postmicrosomal supernatants. The 66-kDa protein appears to be subjected to limited proteolysis, giving rise to a 62-kDa fragment lacking the C terminus and to other more minor fragments of lower M r in the milk-fat-globule membrane. Antipeptide antibodies to epitopes within the N- and C-terminal domains were used to show that butyrophilin retains a type I topology in plasma membranes when expressed in insect cells from a baculovirus vector, and in secreted milk-fat globules. These data do not agree with previous suggestions that butyrophilin may exist in cytoplasmic soluble forms, or be reorganized in the plane of the lipid bilayer

(BTN) (1) with a view to understanding the function of this mammary-specific protein in milk-fat secretion (for reviews of this process, see Refs. [2][3][4]. From the derived amino acid sequence we predicted that BTN is an integral protein with a single membrane anchor and a glycosylated exoplasmic N terminus (type I orientation) (1). Subsequent comparisons with more recently cloned cDNAs have established that BTN is a member of the immunoglobulin superfamily (5) with two Nterminal immunoglobulin domains, one of the intermediate type (IgI) (6) toward the N terminus and one of the constant C1 type (7) toward the membrane bilayer (8). A large proportion of the C-terminal region comprises the RFP or B30.2 domain, a sequence which is present in a subfamily of zinc-finger proteins (9,10), and a group of recently identified BTN-like genes (11,12).
Several workers have proposed that BTN may function as an integral receptor for cytoplasmic fat droplets and that budding of the droplets at the cell surface is initiated by interactions between the cytoplasmic tail of BTN and other proteins, notably the redox enzyme xanthine oxidase (1,(13)(14)(15)(16). However, the assumption that BTN behaves as an integral protein in mammary epithelial cells has been confounded by several observations. There are persistent reports of "soluble" forms of BTN in postmicrosomal supernatant fractions of lactating mammary tissue. BTN has been identified by immunoblotting techniques associated with cytoplasmic lipid droplets (17), which are the presumed precursors of secreted milk-fat globules (17,18), and with a recently described lipoprotein particle which also contains xanthine oxidase, fatty acid synthase, GTP-binding proteins, and lipids (19). 2 BTN is secreted in association with the MFGM in several forms which include not only the isoelectric variants expected for a heterogeneous Nlinked glycoprotein (22,23), but also smaller forms (22,24,25) which could be the soluble products of alternatively spliced mRNAs lacking the membrane anchor.
Morphological analysis of the structure of MFGM has also yielded ambiguous results. In previous immunocytochemical studies, antibodies prepared against purified bovine or guinea pig BTN predominantly decorate the protein coat (24,26) on the cytoplasmic face of the MFGM (13,14), suggesting that BTN is largely located there possibly as a "peripheral" cytoskeletal-like component (13). Furthermore, selective removal of the * This work was supported by Grants C-RFP91-002 and ANSC-93-078 from the Competitive Grants Program of the Maryland Agricultural Experiment Station, Fogarty International Fellowship 1 FO6TWO1508-01 from the National Institutes of Health, and Grant DCB 9119264 from the National Science Foundation (to I. H. M.). 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 2 In preliminary experiments we also apparently identified a soluble form of BTN in postmicrosomal supernatants by immunoblotting techniques (20). These results could not be repeated when Tween 20 was omitted from the washing buffers used for immunoblotting, and subsequently when a different batch of peroxidase-conjugated second antibody was used to detect bound antibody on the blots, either in the presence or absence of Tween 20. This component may have been the heavy chain of endogenous immunoglobulins identified as a nonspecific band on blots of immunoprecipitates developed with chemiluminescence reagents by other workers (21). lipid bilayer from fat globules by physical agitation leaves a substantial amount of BTN associated with the droplet surface in coat structures (17), which is not consistent with the predicted integral properties of BTN (1). These data parallel the results of freeze-fracture electron microscopy from several laboratories which show that the fracture faces of MFGM are depleted of the intramembranous particles typically associated with integral proteins (27)(28)(29).
The situation has been further complicated by the discovery of at least six BTN-like genes in the major histocompatibility complex which are widely expressed in many tissues, possibly including the mammary gland (11,12). Some of these BTN-like proteins (with amino acid identities to BTN approaching 50%) appear to be synthesized from alternatively spliced mRNAs (11). This raises the possibility that earlier studies employing polyclonal antibody to purified BTN have mistakenly identified other BTN-like proteins from this gene family in mammary tissue.
To resolve these issues, we have reinvestigated the expression of BTN in mammary cells using a combination of immunochemical and molecular approaches with epitope-specific antibodies. The results indicate that BTN is synthesized as an integral membrane protein from a single size class of mRNA and retains a type I orientation following secretion in milk-fat droplets.

Materials
Trypsin type XIII and soybean trypsin inhibitor type 1-S were purchased from Sigma. Moloney murine leukemia virus and avian myeloblastosis virus reverse transcriptase, Taq

Preparation of Milk and Tissue Fractions
Washed fat globules for topological studies were prepared from fresh uncooled milk (pooled from three cows) by a method designed to limit physical damage to the outer surface of the droplets (30). Bovine MFGM was prepared as described (31). Soluble (whey) fractions of bovine milk were obtained by centrifuging 200 ml of the pooled skim milk from three cows at 100,000 ϫ g av for 2 h at 4°C. The residual layer of fat was removed by aspiration and the opalescent subnatant carefully separated from the white pellet of casein. Cell fractions were prepared from mammary glands obtained from lactating cows at slaughter using either a Brinkman PT 10/35 homogenizer (Westbury, NY) or a Sorvall Omnimixer, essentially as described (25).

Preparation of Antibodies
Antipeptide antibodies were prepared with peptides synthesized by Dr. Thomas Kempe (Department of Chemistry and Biochemistry, University of Maryland, College Park). For the exo-1 and cyto-1 antibodies (Fig. 1), antigen was treated with immunogen enhancer (CYTImmune Sciences, Inc., College Park, MD) (32) and mixed 1:1 (v/v) with either RIBI or Freund's complete adjuvant. Antigen/adjuvant mixtures were injected subcutaneously into multiple sites on the back of New Zealand White rabbits. For the exo-2 and cyto-2 antibodies (Fig. 1), antigen was coupled to keyhole limpet hemocyanin (33), mixed with Freund's complete adjuvant, and injected into the thigh lymph nodes and subcutaneously into the back of the neck. Subsequent injections of antigen mixed with either Freund's incomplete adjuvant or Tris-buffered saline (140 mM NaCl in 10 mM Tris/HCl buffer, pH 7.4) were administered in the subscapulum, the thigh muscles, and the ear vein.
Antibody to bovine BTN was prepared by using antigen purified from MFGM by preparative SDS-polyacrylamide gel electrophoresis. The purified protein was coupled to keyhole limpet hemocyanin (33) by standard procedures. With the exception of exo-1, antibody was purified by adsorption to, and elution from, the respective antigen (peptide or protein) covalently coupled to Sepharose-4B, as described by Neira and Mather (25). Exo-1 serum was used without purification.

Electrophoresis and Immunoblotting Techniques
SDS-polyacrylamide gel electrophoresis, two-dimensional gel electrophoresis, and ancillary staining techniques were as described (25, 34, 35). For immunoblots, proteins separated by one-or two-dimensional gel electrophoresis were transferred to nitrocellulose (36) and the blots developed according to the procedures of Bio-Rad. Antibody-peroxidase conjugates were detected by enhanced chemiluminescence (Amersham). In some immunoblots Tween 20 was omitted from the washing buffers.

In Vitro Synthesis and Translation of Butyrophilin mRNA
The full-length cDNA clone BT1 (1) encoding bovine BTN was inserted into Bluescript and sense mRNA was synthesized as described by Zerial et al. (37). The synthesized mRNA was translated in the presence of [ 35 S]methionine (Ͼ1,000 Ci/mmol) (30 Ci), a mixture of 19 unlabeled amino acids (without methionine) and rabbit reticulocyte lysate, using the Promega in vitro translation kit. In some experiments, microsomal membranes from dog pancreas, pretreated with 50 mM EGTA (38) were added at a final concentration of 4.0 A 280 nm units/ml. Mixtures were dissolved in SDS sample buffer and separated by SDS-polyacrylamide gel electrophoresis. Gels were soaked in ENHANCE (Amersham), dried, and exposed to x-ray film.

Treatment of Microsomal Membranes with Proteinase K and Triton X-100
Translation mixtures prepared as described above, either in the presence or absence of microsomal membranes, were treated with proteinase K at a final concentration of 0.5 mg/ml for 1 h on ice. In some experiments, the membranes were permeabilized by the addition of Triton X-100 to a final concentration of 1% (v/v). Proteolysis was stopped by the addition of 12.5 mM phenylmethylsulfonyl fluoride and further incubation for 5 min at 0°C.

Immunoprecipitation of BTN Translated in Vitro
Translation mixtures (25 l) prepared in the presence of microsomal membranes were incubated overnight in an end-over-end stirrer with 20 l of antipeptide antibody or control immunoglobulin at 4°C. The microsomes were washed twice with 200-l aliquots of Tris-buffered saline, recovered by centrifugation at 100,000 ϫ g in a Sorvall RCM120 mini-ultracentrifuge at 4°C, and resuspended in 50 l of Tris-buffered saline. Goat anti-rabbit IgG was added at a final dilution of 1:50 and a final volume of 100 l in Tris-buffered saline. Immunoprecipitated membranes were recovered by centrifugation at 1,000 ϫ g after incubating overnight at 4°C in an end-over-end stirrer.

Preparation of Recombinant Baculovirus, vBT
The sequence of DNA encompassing nucleotides 43 through 1655 of clone BT1 (1) was amplified using the PCR, and primer pairs 5Ј-CCCAGAAAGCTAGCAGATGGCAGTCT-3Ј and 5Ј-TGAGCTGGAGTA-GATCTTAAGGCACC-3Ј to prepare a 1.6-kilobase fragment encoding the entire open reading frame of BTN with NheI and BglII restriction sites at the 5Ј-and 3Ј-ends, respectively. This fragment of DNA, after cloning into the pCRII vector, was directionally subcloned into the baculovirus transfer vector pBlueBac II to obtain plasmid pBTBacII.
Spodoptera frugiperda (Sf9) insect cells were cultured in Hink's TNM-FH-medium (39) and cells were supplemented with 10% fetal calf serum, 50 g/ml gentamicin sulfate, and 2.5 g/ml amphotericin B in either Falconware tissue culture flasks at 28°C or spinner flasks (Bellco) at room temperature. Cells grown in spinner flasks contained 0.1% Pluronic F-68 in addition to the above supplemented medium. A recombinant baculovirus, with the BTN cDNA under the control of the polyhedrin promoter, was generated by homologous recombination as described (40). Briefly, 2 ϫ 10 6 Sf9 cells were cotransfected with 2 g of pBTBacII DNA and 1 g of wild-type Autographa californica nuclear polyhedrosis virus DNA. Recombinant virus was isolated by blue/white selection (40) and plaque purified three times. The presence of the BTN cDNA was confirmed by PCR and the recombinant virus designated vBT.

Determination of the Orientation of BTN in Sf9 Cells
Sf9 cells were grown on microscope coverslips in Hink's medium (39) supplemented with 5% (v/v) fetal calf serum. Individual wells of 12-well tissue culture plates were seeded with 0.9 ϫ 10 6 Sf9 cells in 2 ml of medium. After 24 h growth at room temperature, the cells were infected with 1.33 ϫ 10 7 plaque-forming units of vBT. Twenty-four hours postinfection, intact cells on coverslips were washed twice for 5 min each time in Hink's medium containing 0.1% (w/v) NaN 3 at 30°C and then incubated with a 1-50-fold dilution of affinity-purified antipeptide antibody in the same medium for 30 min at room temperature in a humidity chamber. Cells were washed three times with Hink's/NaN 3 medium, for 5 min each time at room temperature, and then fixed for 15 min in 3% (w/v) paraformaldehyde in phosphate-buffered saline (PBS) (140 mM NaCl in 10 mM KH 2 PO 4 /KOH buffer, pH 7.4), containing 0.1 mM CaCl 2 and 0.1 mM MgCl 2 . Reactive aldehyde groups were blocked by incubating the fixed cells for 10 min in NH 4 Cl and the coverslips washed three times, for 5 min each time in PBS. The cells were then incubated with goat anti-rabbit IgG conjugated with fluorescein at a dilution of 1-50fold in PBS for 30 min, washed three times in PBS, rinsed in distilled water, and mounted in elvanol.
To permeabilize cells, coverslips were rinsed three times for 1 min each time in PBS and then incubated with 0.05% (w/v) saponin in PBS for 5 min at 30°C. The cells were fixed in 3% (w/v) paraformaldehyde in PBS, pH 7.4, containing 0.05% (w/v) saponin, 0.1 mM CaCl 2 and 0.1 mM MgCl 2 for 5 min, and then for a further 10 min in the absence of saponin. After a 10-min incubation with NH 4 Cl, the cells were washed three times for 5 min each time in PBS and then labeled with affinitypurified antipeptide antibody in PBS and antibody/fluorescein conjugates in PBS, essentially as described above.

Analysis of mRNA with the Reverse Transcriptase/Polymerase Chain Reaction (RT-PCR)
Total RNA was prepared from lactating bovine mammary tissue obtained at slaughter from the USDA Beltsville abattoir (Beltsville, MD) or from pooled mammary tissue from Balb/c mice (3 mice from day 1, or 3 mice from day 14 of lactation). The RNA was reverse transcribed into cDNA by incubation with either Moloney murine leukemia virus or avian myeloblastosis virus reverse transcriptase and random hexamers at 37 or 42°C, respectively, for 15 min, following the protocol described in the Perkin-Elmer RT-PCR kit (Perkin-Elmer Corp., Branchburg, NJ). Specific regions of the synthesized cDNA, indicated by Arabic numbers in Fig. 2, were amplified by the PCR using Taq DNA polymerase and the primer pairs listed in the legend to Fig. 2. The amplified products obtained with primer pairs 5 (bovine) and 10 (mouse) were sequenced to confirm identity (see also Ref. 41).

Topological Analysis of BTN in Fat Globules and MFGM
Washed cream and MFGM samples were suspended in PBS in 0.5-ml aliquots in 1.5-ml Eppendorf tubes with or without trypsin. Each mixture contained 1.0 mg of protein and 0 -20 g of trypsin. Samples were incubated at 35°C for 30 min and proteolysis was stopped by the addition of 250 g of soybean trypsin inhibitor in 100 l of PBS on ice. The cream and MFGM samples were washed once with PBS containing trypsin inhibitor (1 mg/ml). Fat globules were separated at 5°C by centrifuging in a bench-top Eppendorf microcentrifuge set at maximum speed for 5 min. The bottom of the tube was then cut with a razor blade and the subnatant carefully removed from the packed cream layer with a micropipette. Cream was quantitatively recovered from the tube by centrifuging the contents into a second 1.5-ml Eppendorf tube contained within a 15-ml centrifuge tube. MFGM was recovered by centrifugation at 100,000 ϫ g in a Sorvall RCM120 mini-ultracentrifuge. Cream and MFGM samples were prepared for SDS-polyacrylamide gel electrophoresis and immunoblotting by heating in SDS sample buffer at 95°C for 3 min.

Protein Determination
Protein was assayed, either with the bicinchoninic acid reagent (42), or by the method of Lowry et al. (43) using bovine serum albumin as a standard. Soluble whey fractions from milk were dialyzed against 62.5 mM Tris/HCl buffer, pH 6.8, overnight to remove lactose before protein analysis.

RESULTS
We first investigated the expression of BTN in mammary cells by RT-PCR to establish whether BTN is synthesized from a single sized mRNA or from alternatively spliced mRNAs. Total RNA from the mammary tissue of one lactating cow and two samples of pooled RNA each from three Balb/c mice were reverse-transcribed into cDNA and amplified using the PCR. Primer pairs for the PCR were selected to amplify BTN cDNA (mRNA) across all exon/intron boundaries (41,44) paying special attention to the exon encoding the membrane anchor (exon 4). In every case and in both species, single major amplified products were obtained of the sizes predicted from the published cDNAs (1, 16, 41), indicating that BTN is synthesized from one major size class of mRNA as originally suggested from Northern analysis of bovine RNA (1) (Fig. 2).
To study the distribution and orientation of BTN in cells and tissues, four antipeptide antibodies were produced, two specific for sequences in the N-terminal, exoplasmic domain (exo-1 and exo-2), and two to sequences in the C-terminal, cytoplasmic domain (cyto-1 and cyto-2) (Fig. 1). The antipeptide antibodies were determined to be monospecific by blotting to MFGM proteins that had been separated by SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes. All four antibodies bound to the major band of BTN protein of M r 66,000 on the blots and binding was progressively inhibited by adding increasing amounts of the respective peptide (data not shown). No cross-reactive proteins were detected in microsomal membrane and postmicrosomal supernatant fractions from bovine liver, lung, or heart, with either the exo-2 or cyto-2 antibodies. Similar results were obtained with an affinity purified preparation of rabbit antibody prepared against BTN purified from bovine MFGM (data not shown).
The distribution of BTN in milk and tissue fractions from lactating mammary gland was examined further by immunoblotting techniques. A 66-kDa immunoreactive protein was detected in the microsomal membrane fractions from three cows (example from one shown in Fig. 3, a and b, lanes 3) and in MFGM (Fig. 3, a and b, lanes 1). The soluble postmicrosomal supernatant fractions were without reaction 2 (Fig. 3, a and b,  lanes 4) and only trace amounts of a cross-reactive protein were detected in the soluble whey fraction of milk (Fig. 3, a and b,  lanes 2). Higher M r forms of BTN were also detected in MFGM which appeared as a smear of immunoreactive material toward the top of the blots (bracket, Fig. 3b, lane 1). This material consists of aggregates of the 66-kDa monomer (see Ref. 25 for discussion) which could be resolved into a series of discrete bands when the sensitivity of the chemiluminescent detection method was reduced by omitting Tween 20 from the washing buffers (Fig. 3b, lane 1Ј).
The forms of BTN in MFGM and microsomal membranes were further analyzed by two-dimensional gel electrophoresis (Fig. 3, c and d). Four major isoelectric variants of approximate M r 66,000 with pI values close to the ovalbumin standard (pI 5.1) were detected in the microsomal membrane fraction (Fig.  3, c and d). These same variants were also the major forms of BTN in MFGM detectable with antibody to the whole protein (Fig. 4a). Besides the aggregated material of high M r (Fig. 4a), several other immunoreactive components were detected in MFGM with similar or more basic pI values (numbered 1-5 in Fig. 4a). Epitope mapping with the four antipeptide antibodies showed that variants labeled 2 in Fig. 4a lack the C terminus, since they bound all antibodies except cyto-2 (data not shown and Fig. 4b). Sequencing by Edman degradation confirmed that these variants have an intact N terminus. 3 The identities of other immunoreactive proteins (labeled 3-5 in Fig. 4a) are summarized in Table I. These data indicate that BTN is expressed primarily as a series of acidic variants of approximately 66 kDa in lactating mammary tissue, and that it is subjected to limited proteolysis during or following secretion in association with MFGM. We next investigated the association of BTN with cellular membranes to establish whether the membrane anchor predicted from the cDNA sequence is functional, and to determine the orientation of BTN in cellular membranes and in secreted fat globules. BTN mRNA was synthesized in vitro from the cDNA (1) and the mRNA translated in the presence or absence of dog pancreas microsomes. A 35 S-labeled protein of M r 59,000 was synthesized in the absence of membranes (Fig. 5, lane 1), which is very close to the predicted M r of 59,260 for the pre- . Primer pairs used for each reaction were as follows: bovine, 1, forward 5Ј-CTGAAGTTCCCGACAAACTCG-3Ј; reverse 5Ј-AGCTACTGATGTGAGGATCAGAGCC-3Ј; 2, forward 5Ј-GGCTCTGATCCTCACATCAGT-3Ј, reverse 5Ј-CACTGCATGCAATGTAGCCCT-3Ј; 3, forward, as 2 above, reverse 5Ј-GGACAGTGGGATCTCCTTTGA-3Ј; 4, forward 5Ј-TCAAAG-GAGATCCCACTGTCC-3Ј; reverse 5Ј-CACTGCCTGAGTTCACCTCAT-3Ј; 5, forward 5Ј-GTGGTGGGTGAAGATGCAGAG-3Ј; reverse, as 2 above; 6, forward as 5 above, reverse 5Ј-GGGTCCAGCCAGTGGAGA-3Ј; 7, forward 5Ј-TCCCGAGAATGGGTTCTG-3Ј; reverse, as 4 above; 8, forward, as 4 above; reverse 5Ј-CATTGACCCTGGAAAACCCA-3Ј; mouse, 9, forward 5Ј-CCCTCACTGTCCTACAGCTGCCC-3Ј; reverse 5Ј-TCTCTACTTCCT-TCCCCTGG-3Ј; 10, forward 5Ј-AGACCAGGGGGAGTACCG-3Ј; reverse, 5Ј-CAGGATCTGACGTGAAT-3Ј; 11, forward 5Ј-CCAGGGGAAGGAAG-TAGAGA-3Ј; reverse 5Ј-GTCAAAGGAAGGCAAAGTCG-3Ј. protein estimated from the cDNA sequence. Translation of BTN in the presence of microsomes caused a shift in M r to 62,500 which is close to the size expected for the protein after incorporation into microsomal membranes, followed by removal of a 2.8-kDa signal sequence and addition of two ϳ2.5-kDa N-linked oligosaccharides (Fig. 5, lane 2). To establish that the synthesized protein had associated with microsomes via the predicted membrane anchor (1), samples of BTN synthesized in the presence or absence of sealed microsomes were digested with proteinase K. In the presence of membranes, a 36.5-kDa fragment of BTN was protected from proteolysis (Fig. 5, lanes 3  and 4). Addition of Triton X-100 to permeabilize the membranes in the presence of proteinase K led to complete digestion of the protected fragment (Fig. 5, lanes 5 and 6). This result is exactly as expected if BTN associates with microsomal membranes via the hydrophobic region of 27 contiguous amino acids in the middle of the protein sequence.
The topology of BTN in microsomes was established using the domain-specific antipeptide antibodies. Immunoprecipi-tates recovered with cyto-2 antibody were enriched in membranes containing the 62.5-kDa form of BTN (Fig. 5, inset, lane  2, double arrowheads). This form was not recovered with exo-2 antibody (Fig. 5, inset, lane 1), which is as expected if the protein is incorporated into microsomes with a type I orientation. Both samples contained the 59-kDa unglycosylated form of BTN (Fig. 5, inset, lanes 1 and 2, single arrowhead), which we presume is untranslocated protein, precipitated with either antibody. No labeled protein or vesicles were recovered in the absence of antibody (Fig. 5, inset, lane 3). These data therefore confirm that BTN is incorporated into microsomal vesicles with the N terminus facing the exoplasmic space.
That BTN is also associated with plasma membranes in type I orientation was established using a recombinant baculovirus, vBT, to express BTN in insect cells. Sf9 cells were infected with vBT and 24 h post-infection, labeled with the exo-2 and cyto-2 antibodies before or after permeabilizing the plasma membrane with saponin. As expected for a type I glycoprotein the surface was readily labeled with exo-2 antibody in either intact or permeabilized cells (Fig. 6, a and b). In contrast, the cyto-2 antibody only bound to the cells following permeabilization with detergent (Fig. 6, c and d).
Finally, we determined the topology of BTN in MFGM and secreted milk-fat globules. Fat droplets were removed from bovine milk by a mild flotation procedure (30) and treated with graded amounts of trypsin. Proteolysis was stopped by the addition of trypsin inhibitor and the membrane proteins were analyzed by SDS-polyacrylamide gel electrophoresis and immunoblot (Fig. 7, a-d). Under these conditions, BTN was cleaved from the N terminus into fragments of M r 40,000 -55,000 (Fig. 7, a and c, brackets) that remained associated with the lipid bilayer and that bound to antipeptide antibodies specific for either the C terminus (cyto-2) or to epitopes within the B30.2 domain (cyto-1) (Fig. 7d, brackets). The exo-2 antibody to epitopes within the constant type C1 immunoglobulin domain (Fig. 7e) also bound to the 53-55-kDa fragments. Exo-1 antibody (Fig. 7e) only recognized undigested BTN (Fig. 7d).
The pattern of peptide fragmentation is consistent with cleavage between the IgI and constant type C1 immunoglobulin domain and partial digestion of the remaining exoplasmic sequence. This occurred in the absence of any significant penetration of the lipid bilayer by the proteinase, since xanthine oxidase which is localized on the cytoplasmic face (14,45,46) remained intact during trypsin digestion (Fig. 7b). In contrast, digestion of isolated MFGM under identical conditions led to the cleavage of xanthine oxidase into a 130-kDa fragment and extensive fragmentation of BTN (Fig. 7, b and c). This is a stringent test of the orientation of BTN in fat droplets since the C-terminal sequence of BTN contains 33 trypsin-sensitive sites, including 11 within the first 30 residues distal to the membrane anchor (Fig. 1).
That the N terminus of BTN is exposed on the exoplasmic face of the MFGM was confirmed by digesting intact fat droplets with neuraminidase. Terminal sialic acid residues were cleaved from glycans in the N-terminal region (23) as shown by a slight shift in the electrophoretic mobility of BTN and the production of more basic isoelectric variants after enzyme treatment. Similar results were obtained with isolated MFGM (data not shown).

DISCUSSION
The above results clearly show that BTN is synthesized from a single size class of mRNA in mammary tissue and is incorporated into membranes as an integral glycoprotein of type I orientation in a number of contexts, including dog pancreas microsomes, the plasma membrane of insect cells, and the membrane of secreted milk-fat droplets. BTN is therefore a classical member of the immunoglobulin superfamily (7) with a single glycosylated exoplasmic domain comprising two immunoglobulin-like folds (5,8) and, a functional membrane anchor approximately in the middle of the sequence. Smaller M r forms of BTN in MFGM are proteolytic fragments either produced in situ or during isolation of MFGM in vitro. The most abundant of these forms which comprise 62-kDa fragments of BTN lacking the C terminus (labeled 2 in Fig. 4a) are the ones most consistently identified in other laboratories (see, e.g. Refs. 4,22,and 24). Whether these forms have functional significance is unknown.
The results also confirm that the major immunoreactive proteins labeled 1 and 2 in Fig. 4a are not bovine homologs of the recently discovered family of human BTN-like proteins (11,12) since all the variants labeled 1 reacted with all four anti-  Table I for a summary of immunoreactivities with all four antipeptide antibodies.  Fig. 4. Aggregates of higher M r than monomeric BTN are not included in the analysis. Identification of BTN fragments is based on antibody cross-reactivities and the sizes of the BTN peptides determined by comparison with reference proteins (see Fig. 4  peptide antibodies to epitopes within the exoplasmic and cytoplasmic domains of BTN, and three of the four antibodies reacted with the variants, labeled 2 in Fig. 4a (Table I). Furthermore, all peptide sequences obtained by Edman degradation of the 66-and 62-kDa protein bands are within the sequence derived from bovine BTN cDNA (Refs. 1 and 23, and this study). We cannot exclude the possibility that other BTNlike proteins are associated with the MFGM. However, such proteins will at most comprise a minor fraction of the total membrane-associated protein.
These conclusions are in agreement with some, but not all previous biochemical and morphological studies. The results are at variance with some freeze-fracture studies which showed that MFGM and areas of the apical plasma membrane in close apposition with budding fat droplets are depleted of the intramembranous particles associated with integral membrane proteins (27)(28)(29). Since BTN comprises 20 -40% of the protein content of bovine MFGM (47,48), as an integral protein, it would be expected to give rise to identifiable particles on the fractured membrane halves. The apparent lack of intramembranous particles was taken as evidence that the protein components of MFGM including BTN undergo extensive rearrangement and exclusion from the membrane lipid bilayer during the budding and secretion of fat droplets (27,28). However, in some studies, intra-membranous particles have been identified in freeze-fractured preparations of MFGM, albeit sometimes at lower densities and in clusters, compared with the apical plasma membrane (29,49). The reason for these inconsistencies is unclear.
The results also do not support a previous suggestion that BTN is a component in toto of the cytoplasmic coat of MFGM (24). This conclusion was based on immunocytochemical data (13) and on the observation that BTN is enriched in MFGM coat preparations that are depleted of phospholipid and other membrane proteins (24). It was reasoned that if BTN remains associated with morphologically identifiable coat structures after removal of the membrane bilayer, then BTN must be entirely incorporated into the coat either as a cytoskeletal-type protein or, as noted above, because of a rearrangement of an integral form of BTN during lipid secretion (13). Since BTN retains a type I orientation following secretion (Fig. 7), it is most likely that earlier coat preparations consisted of a stable complex of BTN and detergent-resistant proteins denuded of membrane phospholipid. The structure and constitution of the protein complex intrinsic to the MFGM coat is unclear. Regions of the coat are organized in a highly ordered crystalline lattice with hexagonal symmetry in the milk-fat droplets of some species (49 -51). Because the major coat components are the integral protein BTN (1,24) and the peripheral proteins, xanthine oxidase (14,24,52) and adipocyte differentiation-related protein (53), the protein lattice may be composed of the cytoplasmic tail of BTN cross-linked in a supramolecular complex with xanthine oxidase and adipocyte differentiation-related protein. A recent study has shown that a recombinant fusion protein of the C-terminal domain of mouse BTN specifically associates with xanthine oxidase from mammary epithelial cells cultured in the presence of lactogenic hormones (16). Since xanthine oxidase is a homodimer (54), divalent binding between xanthine oxidase dimers and the cytoplasmic tails of BTN molecules may serve to stabilize coat protein interactions. Associations between the cytoplasmic tail of BTN, xanthine oxidase, and fat globules must be substantial since both proteins remain associated with the fat globule surface following physical removal of the bilayer (17). We are currently testing the hypothesis that BTN and associated proteins form a hydrophobic complex which binds to the surface of cytoplasmic fat droplets and drives expulsion of lipid from the cell.  (14,300). Autoradiogram was exposed for 2 days at Ϫ70°C. Inset, immunoprecipitation of BTN translated in the presence of dog pancreas microsomes. Membranes were precipitated with exo-2 antibody (lane 1), cyto-2 antibody (lane 2), or control serum (lane 3). Samples were separated by SDS-polyacrylamide gel electrophoresis in a 8% (w/v) gel. Autoradiogram was exposed for 14 days at Ϫ80°C. Single and double arrowheads indicate the 59-and 62.5-kDa proteins, respectively, discussed in the text. Our data do not rule out the possibility that some BTN associates as an integral protein with lipid droplets during their formation in the endoplasmic reticulum (17), or with lipoprotein particles containing fatty acid synthase (19) during the subsequent accretion of lipid droplets in the cytoplasm (17)(18)(19). However, these sources are unlikely to contribute significant quantities of BTN to the protein coat of MFGM because BTN comprises a minor fraction of the total protein associated with either lipoprotein particles or cytoplasmic lipid droplets (17,19) and it is not detectable by sensitive immunoblotting techniques in postmicrosomal supernatants (Fig. 3, a and b,  this study).
Demonstration that BTN retains a type I orientation following secretion in association with milk-fat globules is an essen-tial step toward understanding the possible role of BTN in milk-fat secretion and determining the molecular architecture of the MFGM and its associated protein coat. Furthermore, the topology of BTN should allow the selective proteolytic release of the exoplasmic immunoglobulin domains from intact washed fat globules for structural and functional studies.