We thank Jeff Leiden, Craig Thompson, and David Ginsburg for critical review of this manuscript; Brian Seed for the gift of the plasmid pCDM7 and for useful ...
Proc. Natl. Acad. Sci. USA Vol. 86, pp. 8227-8231, November 1989 Biochemistry
Isolation of a cDNA encoding a murine UDPgalactose:1B-D-galactosyl1,4-N-acetyl-D-glucosaminide at-1,3-galactosyltransferase: Expression cloning by gene transfer (oligosaccharide biosynthesis/glycosyltransferase/surface antigen/cloning strategy)
ROBERT D. LARSEN*, VALANILA P. RAJAN*, MELISSA M. RUFF*, JOLANTA KUKOWSKA-LATALLO*, RICHARD D. CUMMINGSt, AND JOHN B. LOWE*t§ *Howard Hughes Medical Institute, and tDepartment of Pathology, University of Michigan Medical Center, Ann Arbor, MI 48109-0650; and tDepartment of Biochemistry, University of Georgia, Athens, GA 30602
Communicated by Phillips W. Robbins, July 24, 1989
We have developed a genetic approach to ABSTRACT isolate cloned cDNA sequences that determine expression of cell surface oligosaccharide structures and their cognate glycosyltransferases. A cDNA library was constructed in a mammalian expression vector by using mRNA from a murine cell line known to express a UDPgalactose:,B-D-galactosyl-1,4-N-acetylD-glucosaminide a-1,3-galactosyltransferase [(al-3)GT; EC 2.4.1.151]. This library was transfected into COS-1 cells, which lack expression of (al-3)GT. Transfected cells containing functional (al-3)GT cDNAs were detected and isolated with a lectin that recognizes the surface-expressed glycoconjugate product of the (al-3)GT enzyme. One cloned (al-3)GT cDNA was rescued from lectin-positive transfected cells. This cDNA contains a single long open reading frame that predicts a 394-amino-acid protein. No significant primary structure similarities were identified between this protein and other known sequences. However, the protein predicts a type II transmembrane topology similar to two other mammalian glycosyltransferases. This topology places the large COOH-terminal domain within the Golgi lumen; this domain was shown to be catalytically active when expressed in COS-1 cells as a portion of a secreted protein A fusion peptide. Biochemical analysis confirmed that this enzyme catalyzes a transglycosylation reaction between UDP-Gal and Gal(fi1-4)GlcNAc to form Gal(al3)Gal(Ji1-4)GlcNAc. This cloning approach may be generally applicable to the isolation of cDNAs encoding other mammalian glycosyltransferases.
antibodies have been successfully used to isolate just two mammalian glycosyltransferase cDNAs (3-6). These more conventional approaches to cloning glycosyltransferase genes face major obstacles since these enzymes are found in small quantities in animal cells and are generally difficult to purify. To circumvent these difficulties, we have developed gene transfer approaches designed to isolate cloned mammalian glycosyltransferase genes and cDNAs (7, 8). These approaches take advantage of existing information about substrate and acceptor properties of glycosyltransferases and make use of the numerous antibody and lectin reagents that are specific for the cell surface-expressed oligosaccharide products of these enzymes. Previous observations have indicated that during murine embryogenesis, dynamic changes occur in the expression patterns of cell surface-localized, terminal, nonreducing Gal(al-3)Gal linkages (9). Similar alterations in the expression of this structure and a cognate UDPgalactose:83-
Oligosaccharide molecules on cell membrane proteins and lipids constitute a major fraction of the surface antigens displayed by animal cells. These structures undergo profound changes during mammalian embryogenesis, suggesting that they may play fundamental roles in development and differentiation processes (1). Cell surface oligosaccharide structure is determined largely by the glycosyltransferase enzymes responsible for catalyzing their synthesis. With few exceptions, each oligosaccharide linkage is the product of a single glycosyltransferase, which is in turn thought to be the product of a distinct glycosyltransferase gene (2). Since many dozens of oligosaccharide linkages have been described (2), it may be expected that at least as many different glycosyltransferase genes exist. Cloned glycosyltransferase genes and their cognate cDNAs represent tools to investigate the molecular mechanisms that regulate the expression of oligosaccharide structure during development and differentiation. However, to date standard molecular cloning approaches that require amino acid sequence information or anti-glycosyltransferase
Construction of an F9 Cell cDNA Library. A cDNA library was prepared from poly(A)+ RNA isolated from retinoic acid-differentiated mouse F9 teratocarcinoma cells by using the procedure of Seed and Arruffo (11, 12) and the mammalian expression vector pCDM7. pCDM7 is a progenitor of the vector pCDM8 (13); pCDM7 lacks the polyoma sequences present in pCDM8, but is otherwise virtually identical (Brian Seed, personal communication). The library contained 3 x 106 independent recombinants. Isolation of a Mouse (al-3)GT cDNA Clone. Plasmid DNA was prepared (14) from an amplified portion of the library and was transfected into COS-1 cells by using the DEAE-dextran procedure (15). Forty samples of 5 x 105 COS-1 cells (in 100-mm dishes) were transfected with 50 ,ug of plasmid DNA each. After a 72-hr expression period, the transfected COS-1
D-galactosyl-1,4-N-acetyl-D-glucosaminide a-1,3-galactosyltransferase [(al-3)GT; EC 2.4.1.151] occur in an in vitro model that mimics some aspects of early murine embryogenesis (10). We report here the successful implementation of a mammalian gene transfer approach to isolate a cloned murine cDNA that determines surface expression of Gal(al-3)Gal linkages and that encodes an (al-3)GT.¶
MATERIALS AND METHODS
Abbreviations: (al-3)GT,
UDPgalactose:,4-D-galactosyl-1,4-N-ace-
tyl-D-glucosaminide a-1,3-galactosyltransferase; GS I-B4, Griffonia simplicifolia isolectin I-B4. §To whom correspondence and reprint requests should be addressed at: Howard Hughes Medical Institute, Medical Science Research Building 1, Room 3510, 1150 West Medical Center Drive, Ann Arbor, MI 48109-0650. IThe sequence reported in this paper has been deposited in the GenBank data base (accession no. M26925).
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.
8227
8228
Biochemistry:
Larsen et al.
cell monolayers were harvested and panned (11, 12) on dishes coated with Griffonia simplicifolia isolectin I B4 (GS I-B4). Lectin panning dishes were prepared (12) by using 10,ug of GS I-B4 per ml in phosphate-buffered saline (PBS) containing 0.1 mM Ca2+ and 0.1 mM Mn2 Plasmid DNA molecules were rescued from adherent cells (16) and were transformed into the Escherichia coli host MC1061/P3 (11). Plasmid DNA was prepared (14) from these transformants and was subjected to an additional screening by the same procedure. Sib selection was subsequently used to screen for plasmids that determined expression of GS I-B4 binding activity in COS-1 cells. E. coli transformants containing plasmid molecules rescued from the second screening were plated to yield 16 pools containing between 100 and 5000 colonies each. Plasmid DNAs were prepared from replica plates and were transfected separately into COS-1 cells, and the transfectants were screened by panning on GS I-B4-coated dishes. These experiments indicated that -1 out of 1000 colonies contained cloned cDNAs determining the GS I-B4-binding phenotype. One "active" =1000-colony pool was subdivided into several smaller pools, and these were each tested for GS I-B4-binding activity. Three subsequent rounds of sib selection with sequentially smaller, active pools identified a single plasmid (pCDM7-aGT) that directed expression of GS I-B4 binding activity in COS-1 cells. Flow Cytometry. COS-1 cells transfected with plasmid DNAs were harvested (8) 48-72 hr after transfection. These were stained with either fluorescein isothiocyanate-conjugated GS I-B4 at 10 ,ug per ml in staining media (7) or with fluorescein isothiocyanate-conjugated GS I-B4 that had been previously incubated with 50 mM raffinose. Cells were then subjected to analysis by fluorescence-activated cell sorting, as described previously (7). Northern Blotting and DNA Sequence Analysis. Northern blots (14) were hybridized with radiolabeled (17) pCDM7aGT cDNA insert at 42°C in a hybridization solution previously described (7). DNA sequencing was performed (18) by using oligodeoxynucleotides synthesized according to the sequence within the cDNA insert. Sequence data base searches and analyses were performed with the Sequence Analysis Software Package published by the University of Wisconsin Genetics Computer Group (19). Assay of (ct1-3)GT and Product Characterization. Extracts were prepared from transfected COS-1 cells as described (7, 8). Cell extracts, conditioned medium from transfected cells, or IgG-Sepharose-bound enzyme was assayed for (al-3)GT by a procedure described previously (10). One unit of (al3)GT activity is defined as 1 pmol of Gal transferred to N-acetyllactosamine acceptor per hour. HPLC-purified, radiolabeled oligosaccharide reaction products were subjected to digestion with either a-galactosidase (Sigma, 20 mU) or B3-galactosidase (Sigma, 1 mU) for 1 hr at 37°C in buffers recommended by the manufacturer. Reaction products were then fractionated by HPLC (10). Methylation analysis of reaction product(s) was carried out as described (10). Construction and Analysis of the Protein A-(al-3)GT Fusion Vector. A 1050-base-pair segment of the (al-3)GT cDNA containing the putative catalytic domain was excised from pCDM7-aGT by digestion with EcoRI. This was cloned into the EcoRI site of pPROTA (20) by using a double-stranded linker (5'-ACGGAATTCCGT-3') to maintain the correct reading frame (see Fig. 3), yielding plasmid pPROTA-aGTc. Plasmids pPROTA-aGTc, pCDM7-aGT, and pPROTA were separately transfected into COS-1 cells. After a 72-hr expression period, the media were harvested and subjected to low-speed (300 x g for 8 min) and high-speed (100,000 x g for 1 hr) centrifugations. Supernatants were then adjusted to 0.05% Tween 20 and were incubated batchwise with 100 gli of preequilibrated IgG-Sepharose or Sepharose 6B overnight at
Proc. Natl. Acad. Sci. USA 86 (1989)
4TC. The matrices were then thoroughly washed (20) and used directly in (al-3)GT assays.
RESULTS
.
A Gene Transfer Approach to Isolate Cloned, Functional, (al-3)GT cDNAs. Tissue- and cell-specific expression of surface-localized terminal Gal(al-3)Gal linkages is associated with expression of cognate (al-3)GTs that catalyze a transglycosylation reaction between UDP-Gal and N-acetyllactosamine (21). COS-1 cells construct surface-expressed polylactosamine molecules (22) that can function as an acceptor substrate for (al-3)GT (10) but do not express this enzyme or its surface-localized product (see below). We therefore expected that cloned cDNAs encoding an (al-3)GT would, if expressed in COS-1 cells, generate the surfacelocalized oligosaccharide product of that enzyme [terminal Gal(a1-3)Gal linkages]. Moreover, these particular transfectants could be isolated by virtue of adherence to plates coated with a lectin (GS I-B4) that specifically binds terminal Gal(a13)Gal linkages (23). The transient expression system developed by Seed and Arruffo (11-13) was used for this approach since it provides for the rescue of transfected cDNAs that determine the expression of cell surface molecules on COS-1 cells and allows the facile construction of large cDNA libraries in a mammalian expression vector. Isolation of a Cloned cDNA That Determines Expression of GS I-B4 Binding Activity in Transfected COS-1 Cells. Mouse F9 teratocarcinoma cells express an (al-3)GT, and this enzyme activity increases concomitant with retinoic acidinduced differentiation of these cells (10). We therefore prepared a cDNA expression library from retinoic aciddifferentiated F9 cells and screened this library for cDNAs that determine expression of GS I-B4 binding activity in transfected COS-1 cells. One plasmid (pCDM7-aGT) was isolated that, when transfected into COS-1 cells, determined expression of surface molecules that directed specific adherence of cells to culture dishes coated with GS I-B4 (data not shown). Fluorescence-activated cell sorting analysis confirmed these observations (Fig. 1). COS-1 cells transfected with pCDM7-aGT, but not cells transfected with pCDM7, stained brightly with fluorescein isothiocyanate-conjugated GS I-B4. This staining could be inhibited with raffinose, a hapten for this lectin (10, 23). These observations indicate that pCDM7-aGT determines de novo expression of surfacelocalized molecules recognized by GS I-B4 and thus expression of terminal Gal(al-3)Gal linkages on cell surface oligosaccharides (23). cDNA Sequence Analysis Predicts a Protein with a Transmembrane Topology. The cDNA insert in pCDM7-aGT is 1500 base pairs long and contains a single long open reading frame in the sense orientation with respect to pCDM7 promoter sequences (Fig. 2). Three methionine codons are found within the first 15 codons of this reading frame; we assigned the most proximal of these as the initiator codon, based on Kozak's rules (24) for mammalian translation initiation. This reading frame predicts a protein of 394 amino acids in length, with a molecular mass of 46,472 Da. Hydropathy analysis (25) indicates that this protein has features of a type II transmembrane molecule (26) that is topologically identical to that predicted for two other mammalian glycosyltransferases (5, 6). This topology predicts a 41-amino-acid, cytoplasmically oriented, NH2-terminal segment; a single transmembrane domain consisting of a 19-amino-acid hydrophobic segment flanked by basic residues; and a large (presumably catalytic) COOH-terminal domain that would ultimately be targeted to the lumen of the Golgi. Two potential N-glycosylation sites are present, indicating that this protein, like other glycosyltransferases, may be synthesized as a glycoprotein. This cDNA sequence contains a long 5' untranslated region, with
Biochemistry: Larsen et al.
Proc. Natl. Acad. Sci. USA 86 (1989)
8229
GenBank, Release 60.0) identified no sequences with significant similarity to the (al-3)GT DNA sequence, including the sequences of a murine 3-1,4-galactosyltransferase (6) and a rat a-2,6-sialyltransferase (5). Expression of a Catalytically-Active, Secreted Protein A(al-3)GT Fusion Protein. We wished to confirm that this cDNA encodes an (al-3)GT and to simultaneously exclude the formal possibility that it instead encodes a trans-acting molecule that induces (al-3)GT activity by interaction with an endogenous gene, transcript, or protein. Therefore, sequences corresponding to the putative catalytic domain of this protein were fused in-frame to a secretable form of the IgG binding domain of Staphylococcus aureus protein A in the mammalian expression vector pPROTA (20), yielding the vector pPROTA-aGTc (Fig. 3). This vector was then tested for its ability to express a catalytically active, secreted and soluble protein A-(al-3)GT fusion protein. COS-1 cells transfected with the pCDM7 vector or with the pPROTA vector generated no detectable cell-associated or released (al-3)GT activity. By contrast, extracts prepared from COS-1 cells transfected with pCDM7-aGT or with the pPROTA-aGTc vector contained 4574 and 20,500 total units, respectively, of (al-3)GT activity. Moreover, conditioned media prepared from cells transfected with pCDM7-aGT or pPROTA-aGTc contained soluble (al-3)GT activity amounting to 4155 units or 50,438 units, respectively. Importantly, the released activity generated by pPROTA-aGTc could be specifically bound to a IgG-Sepharose matrix, whereas the released activity generated by pCDM7-aGT did not interact with this affinity adsorbent (Table 1). These results indicate that this cloned cDNA encodes an (al-3)GT and show that information sufficient to generate a catalytically active (al3)GT resides within the 332 amino acids distal to the putative transmembrane segment. Determination of the Structure of the Trisaccharide Product of the (al-3)GT. Exoglycosidase digestion was used to confirm the a-anomeric linkage predicted for the oligosaccharide product generated by the recombinant enzyme. Radiolabeled trisaccharide product was prepared from UDP-[14C]Gal and N-acetyllactosamine by using the IgG-Sepharose-bound enzyme activity generated by pPROTA-aGTc (see Materials and Methods). Digestion of the HPLC-purified trisaccharide
Log Fluorescence FIG. 1. Flow cytometry analysis of cell surface glycoconjugates on transfected COS-1 cells. (Upper) Fluorescence-activated cell
sorting profiles of COS-1 cells transfected with pCDM7-aGT and stained either with fluorescein isothiocyanate-conjugated GS I-B4 (solid line) or with fluorescein isothiocyanate-conjugated GS I-B4 that had been blocked with the hapten raffinose (dotted line). (Lower) Profiles of COS-1 cells transfected with the vector pCDM7 and stained as described above. Log, logarithm.
ATG codons at -90 and -251, suggesting that translational control mechanisms may participate in the regulation of expression of this sequence (27). This is reminiscent of another mammalian glycosyltransferase whose transcript also contains upstream ATG codons (5). The putative NH2terminal end of this protein lacks a characteristic cleavable signal sequence (28) that may exist in one form of a murine
f3-1,4-galactosyltransferase (6).
Searches of the currently available protein and nucleic acid data bases (Protein Identification Resource, Release 21.0 and
-276 CCTTCCCTTGTAGACTCTTCTTGGAATGAGAAGTAC
-240 -120 1
CGATTCTGCTGAAGACCTCGCGCTCTCAGGCTCTGGGAGTTGGAACCCTGTACCTTCCTTTCCTCTGCTGAGCCCTGCCTCCTTAGGCAGGCCAGAGCTCGACAGAACTCGGTTGCTTTG CTGTTTGCTTTGGAGGGAACACAGCTGACGATGAGGCTGACTTTGAACTCAAGAGATCTGCTTACCCCAGTCTCCTGGAATTAAAGGCCTGTACTACATTTGCCTGGACCTAAGATTTTC M
I
T
M
L
Q D L H V N
K
I
S
M
S
R
S
K
S
E
T
S
L
P
S
S
R
S
G
S
Q E
K
I
M
N
V
K
G
1ATGATCACTATGCTTCAAGATCTCCATGTCAACAAGATCTCCATGTCAAGATCCAAGTCAGAAACAAGTCTTCCATCCTCAAGATCTGGATCACAGGAGAAAATAATGAATGTCAAGGG 41 121 81 241 121 361 161
481 201 601 241 721 281 841 321 961 361
K V
I
L
L
M
L
I
V
S
T
V
V
V
V
F
W
E
Y
V
N
R
I
P
E
V
G
E
N
R
W
Q
K
D W W F
P
S W
AAAGTAATCCTGTTGATGCTGATTGTCTCAACCGTGGTTGTCGTGTTTTGGGAATATGTCAACAGAATTCCAGAGGTTGGTGAGAACAGATGGCAGAAGGACTGGTGGTTCCCAAGCTGG F
K
N
G
T
H
S
Y
Q
E
D N V E G
R
R
E
K
G R N G D R
I
E
E
P
Q
L
W
D W F
N
P
K N
R
TTTAAAAATGGGACCCACAGTTATCAAGAAGACAACGTAGAAGGACGGAGAGAAAAGGGTAGAAATGGAGATCGCATTGAAGAGCCTCAGCTATGGGACTGGTTCAATCCAAAGAACCGC P
D V L T V T P W K A P
I
V
W
E
G
T
Y
D T A L
L
E
K
Y
Y
A
T
Q
K
L
T V G
L
T V
F
CCGGATGTTTTGACAGTGACCCCGTGGAAGGCGCCGATTGTGTGGGAAGGCACTTATGACACAGCTCTGCTGGAAAAGTACTACGCCACACAGAAACTCACTGTGGGGCTGACAGTGTTT A V
G
K
Y
I
E
H
Y
L
E
D F
L
E
S
A
D
M
Y
F
M
V
G H
R
V
I
F
Y
V
M I
D
D
T
S
R
M
P
GCTGTGGGAAAGTACATTGAGCATTACTTAGAAGACTTTCTGGAGTCTGCTGACATGTACTTCATGGTTGGCCATCGGGTCATATTTTACGTCATGATAGACGACACCTCCCGGATGCCT V V
H
L
N
P
L
H
S
L
Q V F
E
I
R
S
E
K
R
W
Q D I
S
M
M
R
M
K
T
I
G
E
H
I
L
A
H
I
GTCGTGCACCTGAACCCTCTACATTCCTTACAAGTCTTTGAGATCAGGTCTGAGAAGAGGTGGCAGGATATCAGCATGATGCGCATGAAGACCATTGGGGAGCACATCCTGGCCCACATC Q H E V D F
L
F
C
M D V D Q V F Q D N F G V E
T
L
G
Q L V A Q L Q A W W Y
K
A
S
CAGCACGAGGTCGACTTCCTCTTCTGCATGGACGTGGATCAAGTCTTTCAAGACAACTTCGGGGTGGAAACTCTGGGCCAGCTGGTAGCACAGCTCCAGGCCTGGTGGTACAAGGCCAGT P
E
K
F
T
Y
E
R
R
E
L
S
A
A
Y
I
P
F
G E G
D F
Y
Y
H
A
A
I
F
G
G
T
P
T
H
I
L
N
L
CCCGAGAAGTTCACCTATGAGAGGCGGGAACTGTCGGCCGCGTACATTCCATTCGGAGAGGGGGATTTTTACTACCACGCGGCCATTTTTGGAGGAACGCCTACTCACATTCTCAACCTC T R E C F K G I
L
Q D K K H D I E A Q W H D E S H L N K Y F L F N K P T K
I
L
S
ACCAGGGAGTGCTTTAAGGGGATCCTCCAGGACAAGAAACATGACATAGAAGCCCAGTGGCATGATGAGAGCCACCTCAACAAATACTTCCTTTTCAACAAACCCACTAAAATCCTATCT
P E Y C W D Y Q I G L P S D I K S V K V A W Q T K E Y N L V R N N V * 1081 CCAGAGTATTGCTGGGACTATCAGATAGGCCTGCCTTCAGATATTAAAAGTGTCAAGGTAGCTTGGCAGACAAAAGAGTATAATTTGGTTAGAAATAATGTCTGACTTCAAATTGTGATG 1201 GAAACTTGACACTATTTCTAACCA
FIG. 2. Sequence analysis of the cDNA insert in pCDM7-aGT. The derived protein sequence is shown in the single-letter code. The putative transmembrane segment is underlined. Asparagine residues that represent potential N-glycosylation sites are at positions 83 and 319.
8230
..
Biochemistry: Larsen et al.
SV40
s.p.
cx(1 ,3)-GT ? catalytic domain
Protein A j
,-O
IVS
Proc. Natl. Acad. Sci. USA 86 (1989)
Protein A uc(1,3)-GT P E G N S U GGGIRATTCCGTfRTTCCAGAG EcoRl
C
B
8010]
EcoRI (Xhol)
(EcoRI) ..
A
_
A
..
FIG. 3. Protein A-(al-3)GT fusion vector construction. The vector pPROTA-aGTc consists of DNA sequences corresponding to
amino acids 63-394 of the (al-3)GT protein sequence fused in-frame to the IgG-binding domain of S. aureus protein A. s.p., transin signal peptide sequences; SV40, simian virus 40 early gene promoter sequences. Sequences denoted by o indicate sections of the vector derived from rabbit p-globin sequences, including segments containing an intervening sequence (IVS) and a polyadenylylation signal (An). EcoRl and Xho I restriction sites used to isolate the catalytic domain from pCDM7-aGT, and that were destroyed during the construction, are noted in parentheses. (Inset) The DNA sequence across the fusion junction and its derived amino acid sequence are shown. The EcoRI site derived from the synthetic linker is underlined.
product with a-galactosidase resulted in quantitative release of [14C]Gal, whereas the trisaccharide was completely resistant to 13-galactosidase digestion (data not shown). To confirm that carbon 3 of the galactose in the Nacetyllactosamine acceptor is involved in the glycosidic linkage formed by the recombinant enzyme, we prepared a [3H]Gal-labeled N-acetyllactosamine acceptor (10) and incubated it with IgG-Sepharose-bound protein A-(al-3)GT activity and 1 mM UDP-Gal under the standard (al-3)GT reaction conditions. The trisaccharide product of this reaction was purified and subjected to methylation analysis (15). Radioactive 2,4,6-trimethylgalactose was identified (Fig. 4). Together, these results indicate that the recombinant enzyme can utilize UDP-Gal and N-acetyllactosamine as substrates to construct a trisaccharide product with the structure
Gal(al-3)Gal(,B1-4)GlcNAc .
Northern Blot Analysis. The (a1-3)GT cDNA hybridizes to single 3.6-kilobase transcript in F9 teratocarcinoma cells (Fig. 5). Our DNA sequence analysis of another cloned (al-3)GT cDNA isolated by colony hybridization indicates that the insert in pCDM7-aGT represents the 5' end of this transcript (data not shown). The remaining 2.1 kilobases of this transcript consists of 3' untranslated sequence not rescued by the expression cloning procedure. The specific activity of (al-3)GT in retinoic aciddifferentiated F9 teratocarcinoma cells is approximately 4fold higher than that in untreated F9 cells (10). Northern blot analysis indicates that steady-state levels of the (al-3)GT transcript also increase concomitant with retinoic acidinduced differentiation of F9 teratocarcinoma cells (Fig. 5). These results are similar to those reported in F9 cells with a
Table 1. Affinity chromatography of (al-3)GT activity released by transfected COS-1 cells into conditioned media Sepharose IgG-Sepharose Vector Applied FT Bound Applied FT Bound 1,662 1,404 ND pCDM7-aGT 1,662 1523 ND pPROTA-aGT, 20,175 712 9717 20,175 18,989 ND Conditioned media containing (al-3)GT activities derived from the pCDM7-aGT or pPROTA-aGT, vector were subjected to chromatography on IgG-Sepharose or on Sepharose. Unbound and matrixretained materials were then assayed for (al-3)GT activity. The values represent the units of (al-3)GT activity. ND, no detectable activity; FT, enzyme activity in flow-through fractions.
E
a.
40
01
1.
.f
0
2
4
6
8
10
12
14
cm from orgin
FIG. 4. Methylation analysis of the (al-3)GT product. The radiolabeled (al-3)GT product (-2000 cpm) was methylated and fractionated by thin-layer chromatography (see Results and ref. 15). Arrows denote the migration positions of methylated galactose standards: A, 2,3,4-trimethylgalactose; B, 2,4,6-trimethylgalactose; C, 2,3,4,6-tetramethylgalactose.
,B-1,4-galactosyltransferase (29) and suggest that the dynamic changes in cell surface oligosaccharide structures known to accompany in vitro differentiation of this cell line (10) are associated with significant changes in glycosyltransferase gene expression.
DISCUSSION Mammalian embryogenesis is accompanied by profound changes in the expression of cell surface molecules, including structurally distinct oligosaccharide linkages (1). The molecular mechanisms that regulate these complex and dynamic glycoconjugate expression patterns remain undefined. However, these patterns are probably a function of a parallel complexity in developmentally regulated glycosyltransferase gene expression, since these enzymes are largely responsible for generating cell surface glycoconjugate structural diversity. Study of the molecular basis for these events requires cloned glycosyltransferase gene segments and cDNAs. The availability of these reagents, in conjunction with tissue- or stage-specific control sequences and transgenic animal techniques, provide mechanisms to perturb normal glycoconjugate expression patterns during development. These studies in turn may generate abnormal phenotypes that may be instructive with regard to the function of cell surface glycoconjugates in the developmental process. We describe here a gene transfer approach for the molecular cloning of a murine cDNA that encodes an (al-3)GT whose oligosaccharide product is expressed in a developmentally regulated fashion. This cDNA appears to encode a transmembrane glycoprotein with a topology virtually identical to the topologies predicted for an a-2,6-sialyltransferase (5) and a,-1,4-galactosyltransferase (6). However, we found 0)
9.5 7.5 4.428S 2.4 18S 1.4 -
FIG. 5. Northern blot analysis. (Upper) Both lanes of this mow.
Northern blot contain 15 jig of total RNA prepared from F9 cells (F9) or from F9 cells that had been treated with retinoic acid (RA/F9) (10). The blot was probed with the radiolabeled (al-3)GT cDNA insert. RNA molecular size standards, in kilobases, and the migration positions of rRNA (28S and 18S) are indicated at the left.
(Lower) Ethidium bromide stainRNA
ing of the 28S rRNA band from equalized RNA samples used to prepare the blot.
Biochemistry: Larsen et al. no extensive primary sequence similarities between either of these two enzymes and the one reported here. This is somewhat surprising since these enzymes use either an identical nucleotide sugar substrate [UDP-Gal for 31,4-galactosyltransferase and (al-3)GT] or an identical acceptor oligosaccharide [N-acetyllactosamine for a-2,6sialyltransferase and (al-3)GT] and since they each probably function within the trans-Golgi and trans-Golgi network compartments. These common substrate requirements and suborganellar locations may reflect convergent evolutionary mechanisms. Our expression studies (Table 1) indicate that, like many other glycosyltransferases (2), the murine (al-3)GT exists in a secreted, soluble form. Our preliminary studies suggest that F9 cells also secrete an active form of (al-3)GT (data not shown). Soluble glycosyltransferases are thought to be derived from membrane-bound precursors by the action of proteases that cleave between the catalytic domain and the transmembrane segment (5). Biosynthetic and structural studies will be necessary to confirm that the secreted form of (al-3)GT is generated by this mechanism. The molecular mass of the (al-3)GT predicted by this cDNA is 46,472 Da. Previous studies reported that a purified murine (al-3)GT has a molecular mass of 80,000 Da (30). This discrepancy might be accounted for by N- and 0-linked glycosylation of the enzyme within the murine host cell. However, it is also possible that these two enzymes represent the products of distinct genes and that they differ in size but exhibit similar or even identical substrate requirements. The cloned cDNA described here identifies multiple restriction fragments within the murine genome in Southern blot hybridization experiments (R.D.L. and J.B.L., unpublished data). Molecular analysis of these sequences and their transcript(s) and comparison of the respective catalytic properties of known (al-3)GTs will be necessary to determine if multiple distinct (a1-3)GT genes exist. It should also be noted that a purified bovine (al-3)GT has a molecular mass of =42,000 Da (31). Whether this enzyme is related to the murine enzyme(s) remains to be determined. In conclusion, the approach described here may allow isolation of other cDNAs that determine expression of cell surface glycoconjugates and their cognate glycosyltransferases, without the need to purify the respective enzymes. This system, as implemented with COS-1 cells, should be applicable to a number of other specific glycosyltransferases that determine terminal glycosylation reactions. Mammalian hosts also exist that display various mutant glycosylation phenotypes (32). These might be employed in conjunction with lectin and antibody selection schemes (32) and with alternative vector systems (33, 34) to isolate specific glycosyltransferase cDNAs. Moreover, sib selection methods based upon sensitive glycosyltransferase assays also provide opportunities for general application of this approach. These approaches may be expected to generate additional molecular reagents with which to study the function of oligosaccharides during development. We thank Jeff Leiden, Craig Thompson, and David Ginsburg for critical review of this manuscript; Brian Seed for the gift of the plasmid pCDM7 and for useful advice; Jeff Bonadio for bringing the plasmid pPROTA to our attention; and Dr. David Smith for helpful
Proc. Natl. Acad. Sci. USA 86 (1989)
8231
suggestions. This work was supported by the Howard Hughes Medical Institute and in part by Grants from the National Institutes of Health (DK38482 to J.B.L. and CA37626 to R.D.C.). 1. Feizi, T. (1985) Nature (London) 314, 53-57. 2. Sadler, J. E. (1984) in Biology of Carbohydrates, eds. Ginsburg, V. & Robbins, P. W. (Wiley, New York), Vol. 2, pp. 199-213. 3. Narimatsu, H., Sinha, S., Brew, K., Okayama, H. & Qasba, P. K. (1986) Proc. Natl. Acad. Sci. USA 83, 4720-4724. 4. Appert, H. E., Rutherford, T. J., Tarr, G. E., Weist, J. S., Thomford, N. R. & McCorquodale, D. J. (1986) Biochem. Biophys. Res. Commun. 139, 163-168. 5. Weinstein, J., Lee, E. U., McEntee, K., Lai, P.-H. & Paulson, J. C. (1987) J. Biol. Chem. 262, 17735-17743. 6. Shaper, N. L., Hollis, G. F., Douglas, J. G., Kirsch, I. R. & Shaper, J. H. (1988) J. Biol. Chem. 263, 10420-10428. 7. Ernst, L. K., Rajan, V. P., Larsen, R. D., Ruff, M. M. & Lowe, J. B. (1989) J. Biol. Chem. 264, 3436-3447. 8. Rajan, V. P., Larsen, R. D., Ajmera, S., Ernst, L. K. & Lowe, J. B. (1989) J. Biol. Chem. 264, 11158-11167. 9. Pennington, J. E., Rastan, S., Roelcke, D. & Feizi, T. (1985) J. Embryol. Exp. Morphol. 90, 335-361. 10. Cummings, R. D. & Mattox, S. A. (1988) J. Biol. Chem. 263, 511-519. 11. Seed, B. & Arruffo, A. (1987) Proc. Natl. Acad. Sci. USA 84, 3365-3369. 12. Aruffo, A. & Seed, B. (1987) Proc. Natl. Acad. Sci. USA 84, 8573-8577. 13. Seed, B. (1987) Nature (London) 329, 840-842. 14. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982) Molecular Cloning: A Laboratory Manual (Cold Spring Harbor Lab., Cold Spring Harbor, NY). 15. Davis, L. G., Dibner, M. D. & Battey, J. F. (1986) Methods in Molecular Biology (Elsevier, New York). 16. Hirt, B. (1967) J. Mol. Biol. 26, 365-369. 17. Feinberg, A. P. & Vogelstein, B. (1983) Anal. Biochem. 132, 6-13. 18. Sanger, F., Nicklen, S. & Coulson, A. R. (1977) Proc. Natl. Acad. Sci. USA 74, 5463-5467. 19. Devereux, J., Haeberli, P. & Smithies, 0. (1984) Nucleic Acids Res. 12, 387-395. 20. Sanchez-Lopez, R., Nicholson, R., Gesnel, M.-C., Matrison, L. M. & Breathnach, R. (1988) J. Biol. Chem. 263, 1189211899. 21. Galili, U., Shohet, S. B., Kobrin, E., Stutts, C. L. M. & Macher, B. A. (1988) J. Biol. Chem. 263, 17755-17762. 22. Fukuda, M., Guan, J.-L. & Rose, J. K. (1988) J. Biol. Chem. 263, 5314-5318. 23. Goldstein, I. J. & Poretz, R. D. (1986) The Lectins: Properties, Functions, and Applications, eds. Liener, I. E., Sharon, N. & Goldstein, I. J. (Academic, Orlando, FL). 24. Kozak, M. (1986) Cell 44, 283-292. 25. Kyte, J. & Doolittle, R. F. (1982) J. Mol. Biol. 157, 105-132. 26. Wickner, W. T. & Lodish, H. F. (1985) Science 230, 400-407. 27. Kozak, M. (1984) Nucleic Acids Res. 9, 3873-3893. 28. von Heinje, G. (1983) Eur. J. Biochem. 133, 17-21. 29. Lopez, L. C., Maillet, C. M., Oleszkowicz, K. & Shur, B. D. (1989) Mol. Cell. Biol. 9, 2370-2377. 30. Elices, M. J., Blake, D. A. & Goldstein, 1. J. (1986) J. Biol. Chem. 261, 6064-6072. 31. Blanken, W. M. & van den Eijnden, D. H. (1985) J. Biol. Chem. 260, 12927-12934. 32. Stanley, P. (1984) Annu. Rev. Genet. 18, 525-552. 33. Margolskee, R., Kavathas, P. & Berg, P. (1988) Mol. Cell. Biol. 8, 2837-2847. 34. Delmas, V., Bastien, C., Scherneck, S. & Feunteun, J. (1985) EMBO J. 4, 1279-1286.
