Generation of Chinese Hamster Ovary Cell Glycosylation Mutants by ...

Tm JOURNAL OF BIOIOGICAL Cmmmv

Vol. 269, No. 5, Issue of February 4, pp. 3717-3724, 1994 Printed in U S A

Generation of Chinese Hamster Ovary Cell Glycosylation Mutants by Retroviral Insertional Mutagenesis INTEGRATION INTO A DISCRETE LOCUS GENERATES MUTANTS EXPRESSING HIGH LEVELS OF N-GLYCOLYLNEURAMINIC ACID* (Received for publication, July 20, 1993, and in revised form, October 8, 1993)

S.Catherine HubbardS, Lorraine Walls, H. Earl Ruleyg, and ElaineA. Muchmoren From the Center for Cancer Research, Massachusetts Institute of lkchnology, Cambridge, Massachusetts 02139 and the Department of Medicine, Division of Hematology /Oncology, the San Diego Veterans Administration Medical Center, and the UCSD Cancer Center, University of California, San Diego, California 92161

In principle, retroviruses can be used as insertional mutagens to isolate genes responsible for recessive phenotypes in mammalian cells. The process is particularly attractive since the genesresponsible for virus-induced phenotypes maybe isolated from the sites of provirus integration. However, in practice, conventionalretroviruses areinefficient mutagens, as evidenced by the fact that between 5 x lo6 and 10’ integration events have been required to disrupt single copy genes (1). Difficulties disrupting specific genes reflect the largesize of the genome, potential biasesagainstintegrationinto specific genes, and biological factors (such as low cloning efficiencies of cells placed in selection) that interfere with the recovery of null clones. These difficulties are compounded by the fact that most mammalian cells are diploid; consequently, no gene responsible for a recessive phenotype in cultured cells has been identified following retrovirus insertional mutagenesis. We have developed two strategies to increase the efficiency of insertional mutagenesis in cultured cells. First, we constructed retrovirus gene trap vectors that confer selectable phenotypes when the virus integrates into expressed genes (2-4). The proportion of cells with virus-induced mutations is two to three orders of magnitude higher in cells isolated after gene trap selection than in cells containing unselected proviruses (5). This enrichment has the advantage of reducing the background of null clones from spontaneous mutants and may permit selection for loss of diploid functions through loss of heterozygosity. Second, we constructed Chinesehamster ovary (CHO)’ cell a murine ecotropic retrovirus receptor, lines that stably express allowing the cells to be infected byMoMuLV-based vectors. CHO cells are functionally hemizygous at a number of loci (6); consequently, a single integration event may result in loss of autosomal gene functions. CHO cells have also beenwidely used by somatic cell geneticists to isolate recessive mutations thataffect a variety of biochemical processes. For example, cytotoxic lectins have been used to select mutants with alteredcell surface glycoconjugate composition. During the past two decades over 20 distinct * This research was supported by Grants R29-GM43165(EAM),RO1CA40602, andR01 HG00684 (to H. E. R.) from the United States Public complementation groups have been described and studied (7Health Service and partially supported by Cancer Center Core Grant 9). Thus, lectin selection of glycosylation mutants inCHO cells P30CA14051 (to P. A. Sharp). The costs of publication of this article provides an ideal system for evaluating retrovirus insertional were defrayed inpart by the payment of page charges. This article must “advertisement” in accordancewith 18 mutagenesis. We recently demonstrated the feasibility of disthereforebeherebymarked rupting the autosomal gene encoding UDP-GlcNAc: a-3-~-manU.S.C. Section 1734 solely to indicate this fact. $ Present address:Genzyme Corp., One Kendall Square, Cambridge, MA 02139. * The abbreviations used are: CHO, Chinese hamster ovary; C o d , 4 Present address: Dept. of MicrobiologyandImmunology, Rm. concanavalin A DMB, 1,2-diamino-4,5-methylenedioxybenzene; Endo AA5206 MCN, Vanderbilt University Schoolof Medicine, 116121st Ave. H, endo-P-N-acetylglucosaminidaseH; L-PHA, leukoagglutinatingphySouth, Nashville, TN 37232-2363. tohemagglutinin from Phaseolus vulgaris; LTR, long terminal repeat; 1 Fkcipient of a Veterans Administration Medical Center Career De- MOI, multiplicity of infection; MoMuLV, Moloney murine leukemia vivelopmentAward. To whomcorrespondenceshouldbeaddressed: rus; PCR, polymerase chain reaction;WGA, wheat germ agglutinin; bp, VAMC, San Diego, 3350 La Jolla Village Dr., San Diego, CA92161. Tel.: base paids); kb, kilobase(s); HPLC, high performance liquid chroma619-552-8585(ext. 3356); Fax: 619-552-7485. tography.

Retroviral insertional mutagenesis can both generate somatic cell mutants and pinpoint the genomic locus associated with a mutant phenotype. In the present study, this approach was applied to Chinese hamster ovary cells (CHO) made susceptible to Moloney murine leukemia virus (MoMuLV)infection by stable expression of an ecotropic retrovirus receptor. These CHO cells were infected with a replication incompetent MoMuLV construct with a promoterless hygromycin phosphotransferase (hygro)gene inserted into theU3 region of the long terminal repeat and a second selectable marker, neomycin phosphotransferase (neo),expressed from an internal promoter. CHO clones containing integrated proviruses were selected with hygromycin or -18, and the subset of these with reduced cell surface NeuSAc were then selected with wheat germ agglutinin (WGA). The majority of the resulting clones had a phenotype not previously described for WGA-resistant CHO mutants arising spontaneously or from chemicalmutagenesis: NeuSAc wasalmostcompletely replaced by NeuSGc. We have provisionally termed these clones S A P mutants, for Fialic acid phenotype. Southern analysis of HindIII digested DNA from four S A P mutants revealed that the MoMuLV provirus is present in a 10.4-kilobase (kb) fragment. Probing with a flanking CHO sequence resulted in equivalent hybridization to a 4.6-kb fragment and the 10.4-kb provirus-containing fragment in all four cases, whileuninfected parental cells and nonS A P glycosylation mutants generated in thesame retrovirus insertional mutagenesis experiments yielded only the 4.6-kb fragment. Sequencing of the 3”flanking DNA revealed that each of the four S A P mutants had a unique provirus integration site falling within a 796 bp region of the CHO genome. Thefrequency with which S A P mutants arisesuggests that thismay be apreferred site for retrovirus integration.

3717

3718

Retrouiral Mutagenesis Induces High Neu5Gc Expression

noside ~-1,2-N-acetyglucosaminyltransferase I (GlcNAc trans- ing, for 2 h with appropriately diluted fresh virus stock in 1.5 ml of medium containing 8 pg/ml Polybrene. This medium was then removed ferase I) by retrovirus insertional mutagenesis(5). In the present study we show that a majority of the CHO and incubation was resumed in 10 ml of normal complete medium. One to 2 days later, selection was begun in medium containing either G418 mutants generatedby sequential selection for provirus integra- (1mg powder/ml) or hygromycin B (0.6 m).Because hygromycinB is tion and resistance to WGA have a phenotype similar to that unstable, medium containing this agent was changed every 3 days reported for MDAY-D2 murine lymphoma subclones (lo), but during selection;G418 mediumwas not changed. Seven to 11days after previously unknown for CHO cells: an alterationof sialic acid initiation of G418 or hygromycin B selection, the second phase of seleccomposition from predominantly Neu5Ac to Neu5Gc. Analysis tion was begun by adding 15 p g / d WGA to the G418- or hygromycinof four of these mutants hasrevealed that all contain an inte- containing medium. Three to 5 days later, WGA-resistantcolonies were grated provirus within a 796-bp segment of the CHO genome. picked fromthe plates by trypsinizing from cloningrings and expanded

in nonselective medium. Under the conditions of these experiments, individual plates contained zero to five WGA-resistant clones. To avoid sibling clones, onlyone colony was generally picked per 100-mm dish. Reagents-The following materials were obtained from the sources A preliminary evaluation of the glycosylation phenotypes of the muindicated: Neu5Ac,Neu5Gc,1,2 diamino-4,5-methylenedioxybenzene tants arising from the double selectionwas performed using a panel of (DMB), hexadimethrine bromide (Polybrene),Sigma; Dowex 5OAG1-X2 lectins, analogous to the procedure described by Stanley (15).Cells were (100-200 mesh, hydrogen form) and Dowex 3-X4A(100-200mesh, chlo- seeded into 24-well plates (lo4 celldwell), and the next day they were ride form), Bio-Rad; G418, Life Technologies Inc.; hygromycin B, Boeh- challenged with a panel of lectins: C o d (2.5, 5, and 10 pg/ml), L-PHA ringer Mannheim; wheat germ agglutinin, E-Y Laboratories; r42- (50 pg/ml),lentil lectin (50pg/ml), and WGA(15,30,70, and 150 pg/ml). 3Hlmannose (10-20 Ci/mmol),AmericanRadiolabeledChemicals; At the same time, each clone was challenged with G418 and hygromycin Pronase, Calbiochem; Hybond-N+ charged nylon membranes and de- B to confirm that all were resistant to G418 but that only hygromycinoxycytidine 5'-[a-3zPltriphosphate, triethylammonium salt (3000Ci/ selected clones were resistant to this agent.(Indeed, all the clones were mmol), multiprime kit, Amersham Corp.; deoxyadenosine8'-(u-thio)tri- found to be G418-resistant, and virtually all G418-selected clones were phosphate P S I (12.5 mCi/ml), New England Nuclear; TA cloning TM sensitive to hygromycin.) system version 1.3, Invitrogen; terminal deoxytransferase, Promega; Two days aRer addition of the lectin, the cells in each well were dGW, Pharmacia LKB Biotechnologies, Inc., agarose, FMC BioProdexamined by light microscopy, and their response was scored. As deucts; cDNA probes forrabbit and bovine cytochromeb5,A. Steggles, NE scribed by Stanley, the most common mutant phenotypes seen after Ohio Schoolof Medicine. Allother chemicals were of reagent grade and WGA selection (see the t e x t ) are readily distinguishable using a lectin were purchased from commercialsources. panel test. Compared with the parental cells, Lecl mutants are hyperReference Cells and Preparation of CHO Clones Stably Expressing a sensitive to C o d , resistant to L-PHA and lentil lectin, and moderately MoMuLVReceptor-All cell lines were maintained as adherent cultures resistant to WGA, Lec8 mutants are extremely resistant to WGA, and in a humidified, 5% COz atmosphere. Clones 18.22, and 44, the CHO Lec2 mutants areless resistant to WGA than Lecl cells. However,this lines used as host cellsfor retrovirus infection, werederivedfrom simple procedure might not reveal clones with novel phenotypes, so a CHO-Kl cells (see below). These cells and the WGA-resistant mutant second, more rigorous evaluation of glycosylation phenotype was necclonesderivedfrom them were maintained in Ham'sF-12medium essary. supplemented with 5% fetal calf serum. The "reference" CHO mutant Analysis of Asn-Linked Oligosaccharides-CHO cells were metabolilines used for comparison of glycosylation phenotypes with the newly cally labeled with ~-[2-~Hlmannose, and radiolabeled glycoproteins isolated clones wereobtained from Dr.Pamela Stanley (Albert Einstein were extracted, Pronase-digested, and treated with endo-/3-N-acetylgluCollege of Medicine, Bronx,NY). These reference CHO lines are a pa- cosaminidase H (Endo H), exactly as previously described(16). Endo H rental line, -0-5, and a series of well characterized spontaneous mu- cleaves between the two innermost GlcNAc residues of oligomannose tants derived from it: Lecl (Pro-Lecl.3C), k c 2 (Pro-Lec2.6A), Lec3 and hybrid, but not complex, Asn-linked oligosaccharides, and the re(Pro-Lec3.4B),and Lec8 (Pro-Lec8.3D).Pro-5 and these Pro-&derived sulting resistant glycopeptides and free oligosaccharides were resolved mutants were maintained in a-modified minimal essential medium on a column of Bio-GelP6 (Bio-Rad, -400 mesh, 1x 110 an)eluted in 0.1 supplemented with 10%fetal calf serum. M pyridinium acetate, pH 5. The fraction size was 0.5 ml. An NIH 3T3 gene for an ecotropic receptor for MoMuLV has previSialic acid was removed from some glycopeptide preparations by mild ously been cloned ( 11)and an expression vector containing this cDNA acid hydrolysisin 0.05 M H2S04 asdescribed (16), and the desialylated (pJET) was obtained from J. Cunningham, Harvard Medical School. glycopeptides were rechromatographed on Bio-Gel P6. Note that this CHO-K1cells (obtained from M. Krieger, MIT) were co-transfectedwith acid hydrolysis procedure differs from that used to release sialic acids a 20:l ratio of pJET to pSVgpt (12). Clones expressing xanthine-gua- prior to analysis of their Neu5Gc content. nine phosphoribosyltransferasewere selected in Ham's F-12 medium Analysis of Sialic Acid Composition-A pellet of 1x lo6 washed cells acid, 2 pg/ml was homogenized into 1ml of 20 IILM sodium phosphate buffer, pH 7.5, with 5% fetal calf serum plus 25&mlmycophenolic aminopterin, 250 pg/ml xanthine, 15 p g / d hypoxanthine, and 22.5 using 5-s pulses times 10 in a Polytron (Brinkmann). This homogenate pg/ml glycine. Twenty-eight resistant clones were tested for MoMuLV was centrifuged a t 75 x g for 15 min; the pellet thus obtained consisted infectability, and three of them (Clones 18, 22, and 44)were found to primarily of nuclei and unbroken cells. The supernatant was centrigive apparent MoMuLV titers similar to those obtained with NIH 3T3 fuged a t 100,000x g for 60min.To the resulting pellet 1ml of 2 M acetic cells (1-2 x 10" G418-resistant colony forming unitdml). acid was added and the mixture heated at 80 "C for 3 h. These condiClone 22 was used for most of the studies described in this report. tions were chosen to maximize release of both Neu5Ac and NeuSGc, Metaphase Clone 22 cells contained an average of20chromosomes with or without @acetylation (17-19). The samples were purified by (range in 22 spreads, 18-23; standard deviation 1.2) compared to an application to sequential columns ofDowex-5OAG (hydrogen form), average of 21 chromosomes observedwith Pro-5 cells (13).2The normal Dowex AG3x4A (formate form) equilibrated in 10 m sodium formate, diploid chromosome number in Chinese hamsters is 22 (14). pH 5.5, and elution with 10 ml of 1M formic acid,as described previously MoMuLVZnfection and Selection of Provirus-containing, WGA-resist- (17-19). The samples were evaporated to dryness to remove formic acid ant CHO Clones-The UBHygroTKNeo MoMuLV used to infect CHO and derivatized with DMB, as described (20). HPLC was carried out cells was harvested from the supernatant of virus-producer cells (y-8), using a TSK gel ODs-12OT column eluted in the isocratic mode with a as previously described (5). For each infection experiment, fresh pro- mixture of acetonitrile/methanoI/water(9:7:84,v/v) and the effluent ducer cell supernatant, passed through a 0.2-pm filter, was used as a monitored over a Linear Fluor LC 304 fluorescence detector at an exvirus stock. The MoMuLV titer was estimated from the number of citation wavelength 374, emission wavelength 448, as described (21). G418-resistant clones arising after infection of NIH 3T3 and clone 22 Using sialic acid standards a linear standard curve can be generated CHO cells with stepwise dilutions of stock. As discussed in the text, the with this technique, so that as little as 1 pmol of sialic acid can be actual number of infectious virions may have been underestimated due detected (data not shown). to the apparent inability of most cells containing a single U3Hygr0Analysis of Integration Site-Genomic DNAs from each of seven CHO TKNeo provirus to survive G418 selection. mutants were digested with Hind111 and separated by agarose gel elecCells were infected with U3HygroTKNeo MoMuLVas follows. Onday trophoresis. Gels were fixed and transferred as described (22), with 1,l-z x IO6cells wereseeded100-mm dish. The next day, whenthe cell modifications recommendedfor transfer to Hybondm-N+ charged nynumber had doubled, the cultures were incubated, with frequent rock- lon membrane. Rehybridization to reduce nonspecificbinding was performed using buffer recommended by Amersham COQ. Hybridization with deoxycytidine 5'-[a-32P]triphosphate, triethyl ammonium salt B. H. Cochran and s. C. Hubbard, unpublished results. EXPERIMENTALPROCEDURES

Retroviral Mutagenesis

Induces High NeuSGc Expression

3719

the experimentsdescribed in this study. When NIH 3T3cells are infected with gene trap vectors such I I as U3HygroTKNe0, selection for the promoterless U3 marker T I I Hindlll EcoRl Pstl ECORl EcoRl P5U generates far fewer resistant colonies than selection for the marker expressed from the internal promoter, reflecting the 5.8 Kb small subsetof proviral insertion sites which enable sufficient transcription of the U3 marker(5). A similar pattern was obF I G . 1. Schematicrestriction map of USHygroTKNeo MoMuLV provirue a~ it would appear after integration into the hostge- served in the present study when clone 22 CHO cells were nome. infected with serial dilutionsof U3HygroTKNeo MoMuLV and selected with either G418 or hygromycin B: about 100 times (3000 CUmmol) linked to the ClaI-Xba 439-base pair fragment (fromthe vector pLNHTL, kindly provided by T.Friedman, Universityof Califor- more G418-resistant colonies were generated thanhygromycin nia at San Diego) encompassing the 5' long terminal repeat (LTR) of B-resistant colonies (data not shown). MoMuLV (23),or to the CHO genomicDNA 275 bp downstream and 300 Generation of CHO Glycosylation Mutants by Retroviral Znbp upstream of the provirus insertion in W4-224g was performed. sertional Mutagenesis-MoMuLV receptor-expressing CHO Membranes were incubated overnight at 65 "C, washed in 2 x SSPE, cells were infected with U3HygroTKNeo virus and selected 0.1% (w/v) SDSat mom temperature followed by 1 x SSPE, 0.1% (w/v) SDS at 65"C for 10 min. Autoradiography was done for 2-5 days at first with G418 or hygromycin B and then with WGA as described under "Experimental Procedures." The resultingclones -70"C.Alternately,oligonucleotideprobes were usedlabeledwith adenosine 5'-[y-s2Pltriphosphatetriethylammonium salt (6000 Ci/ were given designations reflecting their origin and selection mmol). Prehybridization, incubation, and washes were performed at history: Wx-yy-zh "_ (or zg), where the prefix W indicates WGA 42 "C. selection, 5 is the experimentnumber, 121 is the parentalclone Analysis of Proviral Flanking Seqwnces-lkvo strategies were used to analyze genomic DNA flanking the provirus insertion sites. First, number, and is the plate from which the mutant clone was genomic DNA from a SAP mutant with one copy of the provirus con- picked. The terminalh or g indicates selection in hygromycin B struct (W4-22-8g) was digested with PstI, size fractionated on a 1% or G418, respectively. WGA-resistant CHO cells are known to arise spontaneously agarose gel, and then digested with EcoRI. A 1.5-kb fragment which contained the LTR of the construct was incubated with deoxyguanine and tend to accumulate with increasing time in culture, so it triphosphate and terminal deoxytransferase for 60 min at 37"Cfolwas important to assure that the frequency of spontaneous lowed "C 70 by for 15 min as described (24). PCR using WGA-resistant mutants was not so high that it would coman anchor poly(C) primer (CACGTCGACCTAGGCGGCCGCGGCmutagenesis. Only early CCCCCCCCCCCCC) and a primer from the U3RU5 region of the Mo- pletelyobscureretrovirus-induced were used, and the frequency MuLV (CCAGTCCTCCGAITGACTGAGTCG)was performed. Ratio of passages of clones 18,22, and 44 the anchor primer to the U3RU5 primerwas 1:25 for PCR. Second,DNA of spontaneous WGA-resistant colonies was determined to be from each of the SAP clones was used as a template, andPCR was done less than 1/106 cells (data not shown). After infection with using the U3RU5 primer (primer 1)and a primer from the genomic U3HygroTKNeo virus at a calculated multiplicityof infection of DNA 3' of the retrovirusinsertion (primer 2), and primers3 and 4,from 1.5 and, starting 2 days later, selection in hygromycin B for 11 the 5'LTR and genomic DNA flanking the 5' insertion of W4-224g, respectively (Fig.8).For bothstrategies, reactions wereheated to 94 "C days, about 1/103 hygromycin B-resistant colonies was resisfor 5 min andthen allowed to proceed through 40 cycles of denaturation tant to subsequent challenge with WGA, a marked enrichment (94 "C for 1 min), primer annealing (55 "C for 1.5 m i d , and primer over the spontaneousbackground. In parallel cultures infected extension (72"C for 1.5 min). Reactions were then incubated for 7 min at a n multiplicity of infection of 0.15, about lA04G418-resistat 72 "C before being placed at 4 "C. PCR products were purified on a ant colonies was resistant to WGA (data not shown). 2% NuSieve agarose gel and subclonedusing a TA cloning kit (25,26). of Parental and Mutant Asn-linked Oligosaccharide Analysis Clones were prepared using acid phenol for sequence analysis (27). Nucleotide sequences of provirus-cellular DNA junctions were deter- Cells Reveals a New Phenotype-WGA-resistant CHO glycosylation mutants arising spontaneously or after chemical mumined by the dideoxy chain termination method, as described (28). tagenesis have been intensively investigated (9, 29, 30). AlRESULTS though the exact mechanism by which WGA kills susceptible CHO Host Cells and MoMuLV Vector-Before initiation of CHO cells is unknown, in vitro studies indicate thatglycoconthis study an appropriate host cell line and retroviral vector jugates with terminal Neu5Ac are bound to the lectin with were necessary. The retroviral vector chosen for these experi- particularly high affinity (31).Although many potential mutaments is a modified MoMuLV construct, U3HygroTKNeo. Its tions should reduce the level of glycoconjugate Neu5Ac and construction and propertiesare described (51, and a restriction thus confer some degree of WGAresistance, most are very rare. map of the provirus generated after its integration into a host The greatmajority of spontaneous orchemically induced WGAgenome is shown in Fig. 1. U3HygroTKNeo contains coding resistant CHO mutants fall into just four complementation sequences for two selectable markers. One, the neo gene, is groups (7). These four mutant phenotypes, which are recessive, expressed from a n internal herpes simplex thymidine kinase are termed Lecl, Lec2, Lec3, and Lec8. Lec3 does not survive promoter, so that any intact integratedprovirus should confer selection in 15 pg/ml WGA, so most spontaneous mutantsarisresistance to G418. The second marker gene, hygro, lacks a ing during the current study would be expected to be Lecl, promoter and is placed in the U3LTR such that afterproviral Lec2, or Lec8. Lecl mutants lack functional GlcNAc transferintegration its coding sequence initiates 30 bp from the 5' ter- ase I expression (291, Lec2 mutants cannot translocate CMPminus of the proviral genome. This type of retrovirus vector has sialic acids acrossthe Golgi membrane (321, and Lec8 mutants been termed a "gene trap" vector since selection for the U3 cannot translocate UDP-galactose (33). As a result, the glycomarker (in this case with hygromycin B) after infection gener- proteins of each of these mutants carriesa distinct, characterates cell clones with proviruses inserted within transcription- istic setof Asn-linked oligosaccharides. ally active genes (3). To analyze theseAsn-linked oligosaccharides, cellular glycoCHO cells are not normally susceptible to infection by the proteins extracted from each mutant were metabolically laecotropic MoMuLV used in thisstudy. As described under "Ex- beled with r3H1mannose, digested with Pronase, treated with perimental Procedures," CHO-K1 cells were made MoMuLV Endo H to release oligomannose oligosaccharides, and analyzed infectable by selecting stable clones (clones 18, 22, and 44) by gel filtration chromatography. The results of this type of expressing functional ecotropic virus receptor cloned from NIH analysis provide a great dealof information on the Asn-linked 3T3 cells. Clone 22 was used as the MoMuLV host for most of oligosaccharide composition of the cells (13)and readily distin-

z

Retroviral Mutagenesis Induces High NeuSGc Expression

3720

2M)L 060

60

3ooo

i

80

120

140

160

B Lec1

800

L

Zoo0

600

-

400 loo0

-

200

0 140

160

60

1 2 0 80

100

140

160

Fraction Number FIG.2. Glycosylation phenotypes of three previously characterizedWGA-resistant CHO mutants and the parental strain from which they were derived. Parental (Pro-5) and three WGA-resistant CHO cell lines obtained from P. Stanley were metabolically labeled with [SHlmannose,and labeled glycoproteins wereextracted and digested to glycopeptides with Pronase. After incubation with Endo H to release free hybrid and oligomannose oligosaccharides, the mixture of radiolabeled complex glycopeptides and oligosaccharides was partially resolved by Bio-Gel P6 gel filtration chromatography. On this column, complex glycopeptideselute between the exclusion volume (Ve)and fraction 120; Endo H-released hybrid oligosaccharides containing 1Neu5Ac residue and 4 (H1) or 5 ( H l a )mannose residues elute between fractions 125 and 130; and oligomannoseoligosaccharides GlcMaQGlcNAc through Man,GlcNAc elute between fractions 130 and 160. M a ,migration position of a Man5GlcNAc standard. The peaks labeled B2,B l , and BO comigrate with biantennary complex glycopeptide standards containing 2, 1, or no Neu5Ac residues, respectively; these and the other standards were obtained from Pronase- and Endo H-digested Sindbis virus grown in chicken embryo fibroblasts (51). Shown are gel filtration profiles for Pro-5 cells(panel A ) and three WGA-resistant mutants derived from this line, Lecl (panel B ) , Lec2 (mmel C ) , and Lec8 (panel D ) . Note that Pro-5 was not the parental CHO line used in this study; the glycosylation phenotype of CHO-Kl cells is ;lightly different.

guish the phenotypes of mutant CHO celllines with previously described WGA-resistant phenotypes (Fig. 2). Therefore, this method was used to analyze all the WGA-resistant mutants arising from retroviral insertional mutagenesis. Expression of the MoMuLV receptor had no effect onAm-linked oligosaccharide composition of CHO cells (data not shown). Many of the WGA-resistant, provirus-containing CHO clones obtained after infection of clone 22 cells with U3HygroTKNeo virus had Asn-linked oligosaccharide compositionsresembling those of previously described WGA-resistant mutants (see below). However, at least half of the provirus-containing WGAresistant clones had a glycosylation pattern unlike that of any previously known CHO mutant. For these mutants, the complex glycopeptide gel filtration profile was shifted slightly toward the inclusion volume compared with glycopeptides from the parental strain (Fig. 3). After removal of sialic acid residues, the complex glycopeptides from these atypical mutants closely resembled those from the parental cells (Fig. 41, suggesting that sialic acid compositionhad been altered. For this reason, the new mutant phenotype has beenprovisionally termed SAP, for sialic gcid phenotype; it will be given a standard Lec (or LEC) designation after it isdetermined to be recessive (or dominant). The phenotypes of 49 independent provirus-containing

WGA-resistant clones and 38 spontaneous WGA-resistant clones are summarized in Table I. Both groups include clones with glycosylation patterns resembling those of known WGAresistant mutants Lecl, Lec2, and k c 8 (gel filtration profiles not shown but virtually identical to those shown in Fig. 2). However, while SAP mutants were by far the most common phenotype observed with provirus-containing WGA-resistant mutants, this phenotype was never observed in the spontaneous mutants. This result, together with the fact that the SAP phenotype has never before been reported for spontaneous or chemically induced WGA-resistant CHO mutants, suggests that the SAP phenotype is very rare in CHO cells except after retroviral insertional mutagenesis. Table I also shows that SAP mutants were the predominant WGA-resistant phenotype of provirus-containing clones after either hygromycin B or G418 selection. Thus, the gene trap strategy was not necessary for detection of this provirus-associated mutation. The Major Sialic Acid in the Membrane Fraction in Parental CHO Cells Is Neu5Ac and Neu5Gc in the SAP Mutants-The sialic acids from the parental cell lines (CHO-K1 and clone 22 and four independent SAP mutants derived from clone 22(W322-2h, W3-224h, W4-22-8g, W5-223g) were released, puritied, and analyzed by HPLC on a TSK-ODS column after de-

3721

Retrovirul Mutagenesis Induces High NeuSGc Expression 1000

I

I

2000

1

I

800

600

400

200

0 60

160

80

140100

120

Fraction Number FIG.3. Comparison of complex glycopeptides from uninfected clone 22 CHO cells and from a MoMuLV provirus-containingSAP mutant. [3Hlmannose-labeled,Pronase- and Endo H-digested glycopeptides fromuninfected clone 22 CHOcells (dashed line with open circles) or from a SAP mutant (heauy linewith closed circles)were subjected to gel filtration chromatography as described in Fig. 2. The results of two runs on the same column are superimposed for comparison;migration positions of exclusion and inclusion volume markers inthese runs were the same. The SAP mutant profile shown (from clone W1-1SM) is indistinguishable from those of dozens of other independent SAP mutant clones, including all of those described in this study(data not shown). The radioactivity values shown on the left side of the plot are for the SAP mutant, while those on the right are from the uninfected clone 22 cells.

Frc. 4. Complex glycopeptides of parental and SAP mutant cells before and after sialic acidremoval. Panel A, [3Hlmannose-labeled, Pronase-digested glycopeptidesfrom uninfected clone22 cells ( f i n e line with open symbols) versus those from a SAP mutant (heauy linewith closed symbols). Panel B , the same two glycopeptide samples after removal of sialic acidby mild acid hydrolysis. For both panels, the two profiles shown are from consecutive runs on the same Bio-Gel P6 gel filtration column. The radioactivity values shown on the left sides of the plots are for the SAP mutant, while those on the rightare from the uninfected Clone 22 cells. Note that after sialic acid removal, both cell lines have glycopeptide profiles resembling that from Lec2 mutant cells, which synthesize N-linked oligosaccharides lacking sialic acid (see Fig. 2C).

so

100

120

FRACTION NUMBER rivatization with DMB, as described under “Experimental Procedures.” The sialic acids released are comprised entirely of unsubstituted Neu5Ac and Neu5Gc. Shown in Fig. 5 are the percentages of membrane-associated Neu5Gc compared with total sialic acids in parent and S A P mutant cells. In this experiment, the level of Neu5Gc in the parent is 8.8%, in W322-2h 90.4%, W3-224h 88.1%,W4-22-8g 85.9%, W5-22-3g

88.5%. The replacement of Neu5GcforNeu5Ac in the SAP mutants is consistent with their WGA-resistant phenotype, since WGA does not bind to Neu5Gc 134). Two each of the Lecl-like, Lec2-like, and Lec8-like mutants were analyzed by the same methods, and these were found to have parental levels of Neu5Gc (data not shown). The degree of sensitivity of the S A P mutants to a panel of cytotoxic lectins with different car-

Retroviral Mutagenesis Induces High Neu5Gc Expression

3722

TABLEI

1

Apparent phenotypes of WGA-resistant mutant CHO clones isolated with or without MoMuLV infection Apparent phenotypes of individual, independent clones were classified according to their lectin resistance and gel filtration chromatography of their Pronase- and Endo H-digested cellular glycoproteins a s described in the text. Note that similarity of glycosylation pattern does not constitute proofof identity with the established WGA-resistant mutants Lecl, Lec2, and Le&. The MoMuLV-infected clones in this table were obtained by infecting clone 22 CHO cells with U3HygroTKNeo MoMuLV (calculated multiplicity of infection range 0.023)and selecting first for proviral intergration (with hygromycin B or G418) and then for WGA resistance a s described in the text. The spontaneously arising WGA-resistant clones were obtained byWGA treatment of uninfected clone 22 or clone 44 CHO cells. Apparent phenotype

Number of WGAR clones analyzed Spontaneous MoMuLV-infected clone 22 Hygmmycin G418 Clone 22 Clone 44

Othera

7 4 4 14 0

Total

29

Lecl kc2 Lec8

SAP

6 14

1 3 1 15 0

2 9 4

2

0

0

1

0

20

16

22

The gel filtration profile from this clone did not resemble those of Lecl, Lec2, Lec8, or SAP mutants. It was not characterized further.

2

3

4

5

6

7

8

9

24.8 \ 22.6 19.4 17.1 15.0C

--

8.6 -

12.2

-1

10.1

0.4 kb

8.3

w

6. Southern analysisofparental cells and mutants probed with MoMuLV LTR DNA from parental cells and WGA-resistant mutants (SAP mutants, Lecl-like, and Lec8-like) isolated after infection with U3HygroTKNeo MoMuLV was digested with HindIII, and 20 pg of each digest loaded. After electrophoresis, DNA was transferred onto HybondTMand probed using the MoMuLVLTR, a s described under "Experimental Procedures." Lanes 1 and 2 show DNA from the CHO parent; lane 3, clone 22; lane 4, W3-22-2h (SAP);lane 5,W3-22-6h (SAP);lane 6,W4-22-8g (SAP);lane 7, W5-22-3g (SAP);lane 8, W522-26hA (Lecl-like);lam 9, W5-22-27h (Lec8-like).At the arrow is the 10.4 kb band common to all SAP mutants. FIG.

1 2 3 4

5 6 7 8

410.4kb

cell

type

F I G . 5. Membranebound sialic acids in CHO-Kl parent and mutant cells. 1 x lo6 cells were homogenized and fractionated by ultracentrifugation (see "Experimental Procedures"). Sialic acids were released with acetic acid and quantified using derivatization with DMB over a TSK-ODs-120T column. The sialic acids are expressed a s percent Neu5Gc of the total sialic acid released. The only other sialic acid present in these specimens is unsubstituted Neu5Ac. Lane 1, CHO-K1 parent; lane 2, clone 22; lane 3, W3-22-2h (SAP);lane 4, W3-33-6h (SAP);lane 5, w4-22-8g (SAP).

bohydrate specificities (L-PHA, lentil lectin, C o d , and WGA) closely parallels that of Lec2 mutants (datanot shown). Since Lec2 CHO cells lack sialic acid on their glycoconjugates but appear to be otherwise normal, this suggests that the SAP phenotype does not encompass gross alterations of glycosylation status except for sialic acid composition. Southern Analysis of Proviral ZntegrationSites-Because there isno HindIII site in theproviral construct (Fig. l),DNA isolated from each of the clones was digested with HindIII, subjected to Southern analysis, and probed with MoMuLV LTR to determine numbers of proviral inserts. As seen in Fig. 6, neither the parent nor clone22 has a genomic area which hybridizes with the LTR fragment. Both W3-22-2h and W322-6h have three proviral inserts, W5-22-3g has two, and W4-22-8g, W5-22-26hA, and W5-22-27h have one. Of note is that all four SAP mutants (W3-22-2h, W3-22-6h, W4-22-8g, and W5-22-3g) analyzed share a fragment ofcommon size (10.4 kb), which the Lecl-like mutant (W5-22-26hA) and the LecSlike mutant (W5-22-27h) do not have, implying that this site of integration is significant for generation of SAP mutants.

4 4.6 kb

FIG. 7. Southern analysis of parental cellband mutants probed with CHO DNA-flanking provirus insertion. DNA from parental cells and WGA-resistant mutants (SAP mutants, Lecl-like, and Lec8like) isolated after infection with U3HygroTKNeo MoMuLV was digested with HindIII and 20 pg of each digest loaded. After electrophoresis, DNA was transferred onto Hybondm and probed with CHO DNA containing sequences upstream and downstream of the W4-22-8g insertion, as described under "Experimental Procedures." Lane 1 shows DNA from the CHO-K1 parent; lane 2, clone 22; lane 3, W2-22-2h (SAP);lane 4, W3-22-6h (SAP);lane 5, W4-22-8g (SAP);lane 6,W52 2 3 g (SAP);lane 7, W5-22-26hA (Lecl-like); lane 8, W5-22-27h (Lec8-like). Shown are the 10.4 kb band common to all SAP mutants seen previously when the probe was MoMuLV LTR and a 4.6-kb fragment common to all cell lines which is thought to represent the wildtype allele.

This Southern blot was stripped and probed with a CHO genomic fragment containing 300 bp upstream and 275 bp downstream of the W4-22-8g insertion (see "Experimental Procedures" for construction of this fragment). As seen in Fig. 7, there ishybridization to the 10.4-kb fragment seen when SAP

Retroviral Mutagenesis Induces High NeuSGC Expression

3723

preferentially generates mutants with a phenotype not observed for WGA-resistant CHO mutants arisingspontaneously or from chemicalmutagenesis. In mutantswith this phenotype, provisionally termed SAP, Neu5Gc, normally present at much Primer 2 lower levels than Neu5Ac, is the major sialic acid in cellular glycoconjugates. Analysis ofCHO sequences flanking integrated proviruses in four S A P mutants revealed that in each w4-22-89 W5-22-39 W3-22-2h W3-22-6h case the integration had occurred within a 796-bp regionof the genome. Thus, provirus insertion into thislocus alters theregu795 nt 4 + lation of sialic acid metabolism. FIG.8. Diagrammaticrepresentation of provirus insertion If the four SAP mutants for which the integration locus is sites in SAP mutants. Using a strategy and primers as described known are assumed to be representative, then an approximate under “Experimental Procedures,”the insertion site of the provirus was calculation can be made to compare the observed frequency of determined for four independent SAP mutants, W3-22-2h, W3-22-6h, W4-2-g. and W5-22-3g. The distance between insertions of W P provirus integration into this 796 bp regionwith the predicted, 22-8g and W3-223g is 297 bp, from WP22-8g to W3-22-2h, 250 bp, assuming integration is completely random. For the 3 x 109-bp and from W3-222h to W3-22-6h, 249 bp. The intervening genomic haploid rodent genome, the fraction of random integration sequence is completely overlapping, without nucleotide substitutions or deletions. Note that theU3HygroTKNeo provirus is not drawn to scale events expected to occur within an 800-bp target would be less than 1 x lo6. However, the proportion of SAP clones among and isincluded onlyto show locations of PCR primers 1 and 3. Primers 1and 3 are drawn showing only relevant hybridization. G418- and hygromycin B-resistant clones was much higher than would be expected given the measured multiplicities of Hind111 digests are probed with the LTR. There is also hybrid- infection. Thus, approximately one SAP mutant was obtained ization to a 4.6-kb fragment which appears in allcell lines. The per lo4 and lo3 G418-resistant and hygro-resistant clones, resimilarity of size between the SAP clones indicates that no spectively. This suggests that the796 bp region mayrepresent major deletions or rearrangements have occurred. A separate a hot spot for retrovirus integration. Indeed other examples of Southern analysis of parental and W4-22-8g EcoRI digests preferred regions of virus integrationhave been described (35, probed with this same fragment showed bands which were of 36). predicted size, again indicating that deletions or rearrangeHowever, the magnitude of the bias for this site isdifficult to menta wereminor, if present a t all (data not shown). The estimate, since the actual number of proviruses in each clone 10.4-kb fragment is exactly the anticipated size of the allele was 5-10 times higher than might be expected from the biowith integrated provirus based on addition of the vector and the logical titer (measured as G418-resistant-transducing unitsl apparent wild-type allele. Densitometric analysis revealed that milliliter of producer cell supernatant). This discrepancy is the 4.6- and 10.4-kb fragments had equal intensity in each of large enough to suggest that expression of the TKNeo gene is the four SAP clones (data not shown), which also suggests that sometimes not sufficient to conferG418 resistance in CHO these fragmentscorrespond to the two alleles. The fact that all cells, probably becausethe thymidine kinase promoter is relaS A P mutants share thesetwo fragments implies that CHO-K1 tively weak. As a result, selection for G418resistance may favor cells are either functionally hemizygous in this region or that clones in which the proviruses have inserted in regions favorthe provirus-induced SAP phenotype is dominant. able for thymidine kinase promoter activity. The virus would Proviral Integration Occurs within a 796-bp Segment of the function as an “enhancer trap” andenrich for clonescontaining CHO Genome in Four SAP Mutants-As described under “Ex- proviruses inserted near expressed genes, and titers based on perimental Procedures,” gel-purified PstI fragments ofW4G418 resistance would underestimate the actual multiplicity of 22-8g DNA were digested with EcoRI and amplified by PCR to infection. Nevertheless, from the average number of proviruses determine the CHO sequence downstream of the provirus. This per WGA-resistant clones as measured by Southern blot analmaterial was cloned and sequenced, and a primer was made to ysis (data not shown) each S A P mutant selected required less analyze the other SAP mutants. Using primers 1and 2 (Fig. 81, than lo6 integration events, suggesting at least a &fold prefPCR of the other SAP mutants (W3-22-2h, W3-22-6h, and erence for the site. W5-22-3g) was performed under the same conditions. A single Gene trap selection enriches for cell clonesin which the proPCR product was obtained in all clonesexceptW3-22-6h, virus has integrated into expressed genes (2-5). This has the which had no product (data not shown). The sizes of the frag- advantab? of reducing the spontaneous background of null ments generated varied: 450 bp forW4-22-8g, 750 bp for W5- clones and may permit subsequent selection for loss of diploid 223g, and 150 bp for W3-22-2h. Subcloning and sequence gene functions (through loss of heterozygosity). However,neianalysis of the fragments obtained has revealed the point of ther feature of gene trap selection wasnecessary in the present proviral integration for each of these SAP clones. Subsequently, study since no spontaneous SAP mutants arose in uninfected PCR using primers 3 and 4 (“Experimental Procedures” and cells and a single, common integration event appeared suffiFig. 8) was performed, and this revealed the insertion site for cient to induce the SAP phenotype. Consequently,SAP mutants clone W3-22-6h, whichis downstream of the other insertions. comprise the majority of WGA-resistant clones after eitherhyProviral integration is inan area encompassing 796 bpof CHO gromycin B (gene trap) or G418 selection. genomic DNA for all the clones (Fig. 8). The host genomic seThe 796-bp CHO genomic region associated with the SAP quences within the overlapping PCR clones shownin Fig. 8 are phenotype has no obvious homologywith GenBank sequences, identical for all four SAP clones. There are no apparent open nor doesit contain a clear open reading frame. One interesting reading frames within this region, and this sequence does not feature isa long simple repeat, (dC-dA)zl,near the 3’ end of the have homology with any in GenBank. However, there is a region. Simple purine-pyrimidine sequences have been implisimple repeat of a purine-pyrimidine sequence which is cated as genomic ”hot spots” for recombination (37, 38). It has 30 bases upstream of the W3-22-6h insertion. been determined that (dT-dG),’ (dC-dA), adopts a left-handed helical structure (391, and that it is an abundantsequence in DISCUSSION eukaryotic DNA (40,411. It is therefore of interest to speculate In this report we demonstrate that retrovirus insertional that the(dC-dA)zlwithin the common insertion area may have mutagenesis followed by selection in wheat germ agglutinin served as a recognition site for provirus integration, although Primer 3

c

Primer 1

+

3724

Retroviral Mutagenesis Induces High NeuSG‘c Expression

this has not been observed at other provirus integration sites. Most mutations induced by provirus integration are expected to be recessive, since dominant mutations associated with procultured virus integration (e.g. oncogene activation) are rare in cells. In addition, we used hypodiploidCHO cells and gene trap vectors from whichviral enhancer sequences had been deleted to facilitate the recovery of recessive mutants. Loss of autosomal gene functions was illustrated previously by the isolation of nullizygous cells containing a disrupted GlcNAc transferase I gene (5). LecZ, Lec3, and Lec8 phenotypes are also caused by loss of function mutations. Similarly, the presence of a common integration site in four independent mutants suggests that the disruption of a single gene was responsible for the SAP phenotype. If this reasoning is correct, then this represents the first instance in which a gene responsible for a recessive phenotype in cultured cells has been identified by retrovirus insertional mutagenesis. The SAP mutants contained Hind111 restriction fragments of 10.4 and 4.6 kb, corresponding to the disrupted and normal alleles, respectively. Thus,the mutantswere heterozygouswith regard to the site of virus integration. We assume that the unmutated allele is not expressed, as has been observed fiequently in CHO mutants. However, further experiments are required to exclude the possibility that the phenotype results from a dominant effect of virus integration. The fact that theSAP phenotype affects the Neu5Gc content of cellular glycoconjugates is intriguingbecause this sialic acid is an onco-fetal antigen in humans(42,43), and its expression in multiple tissues of Sprague-Dawley rats has been shown to be developmentally regulated (17,ZO). Adult human tissuesdo not contain this sialic acid, and exposure to Neu5Gc-containing glycolipids has been implicated as a cause of “serum sickness” in patients receiving injections of blood products from animal sources (44,451. The regulation of Neu5Gc expression is thusof considerable interest, butvery little isknown about it. Neu5Gc is synthesized at the nucleotide sugar level by CMP-Neu5Ac hydroxylase (46, 471, and cytochrome b5 is a cofactor in this reaction (48, 49). Neither human nor rabbit cytochrome b5 cDNA hybridized to CHO DNA immediately flanking provirus insertion (data not shown), so that it appears that this gene is not the target.The only other mutant cell line reported to have elevated levels of Neu5Gc is a derivative of the MDAY-DZ murine lymphoma cellline described by Dennis (10).In the latter case, the mutant phenotype could be accounted for simply by increased activity of CMP-Neu5Ac hydroxylase (50). Further work is required to elucidate the mechanism by which provirus integration intoa 796-bp locusalters sialic acid metabolism in SAP mutants. It is likely that some of the mutants with Lecl-, LecZ-, and Lec8-like Asn-linked oligosaccharidecompositions listed in Table I acquired their phenotypes by provirus insertion. These mutants may allow other genes involved in the synthesis of N-linked oligosaccharides to be identified as sequences associated with common sites of provirus integration. Indeed, the use of gene trap retrovirus together with hypodiploid CHO cells is expected to provide a general method for isolating genes responsible for recessive phenotypes in mammalian cells. Acknowledgments-We thank Wen Chang for helpin early molecular analysis of the WGA-resistant CHO mutant cell lines, Pam Stanley for graciously providing Pro-5 and mutant CHO cells, and P. W. Robbins and P. A. Sharp for support. We also thank Celeste Bailey and Yaakov Frishman for helpful discussions regarding PCR strategies.

REFERENCES 1. Varmus, H. E., Quintrell, N., and Ortiz, S . (1981) Cell 26.23-36 2. von Melchner, and Ruley, H. E. (1989)J. Virol. 69,32274233 3. von Melchner, H., Reddy, S., and Ruley, H. E. (1990) Proc. Natl. Acad. Sei. U.S. A 87,3733-3737 4. Chen, J., DeGregori, J., Hicks, G., Roshon, M.,Scherer, C., Shi, E., and Ruley, H. E. Methods Mol. Genetics 4, in press 5. Chang, W., Hubbard, S . C., Friedel, C., and Ruley, H. E. (1993) Wrolwy 19S. 737-747 6. Siminovitch, L. (1985) in Molecular Cell Genetics (Gottesman, M. M., ed) pp. 86W79, Wiley, New York 7. Stanley (1981) Mol. Cell Biol. 1, 687-696 8. Stanley, P., and Chaney, W. (1985) Mol. Cell Bwl. 6, 1204-1211 9. Stanley, P., (1983) Methods Enzymol. 96, 157-184 10. Dennis, J. W. (1986) Cancer Res. 46,4594-4600 11. Albritton, L. M., Tseng, L., Scadden, D., and Cunningbam, J. M. (1989)Cell 57, 659-666 12. Mulligan, R. C., and Berg, P. (1981) Proc. Natl. Acad. Sei. U. S . A 78, 20722076 13. Stanley, P., Caillibot, V,and Siminovitch, L. (1975) Somot. Cell Genet. 1,3-26 14. Stanley, P., Caillibot, V,and Siminovitch, L. (1975) Cell 6, 121-128 15. Stanley, P. (1985) Mol. Cell. Biol. 5, 92%929 16. Hubbard, S.C. (1987) J. Biol. Chem. 262, 16403-16411 17. Muchmore, E. A., Varki, N., Fukuda, M., and Varki, A. (1987) FMEB J . 1, 229-235 18. Varki, A,, and Diaz, S . (1984)Anal. Biochem. 137,236-247 19. Diu, S., and Varki, A. (1985) Anal. Biochem. laO,32-46 20. Muchmore, E. A. (1992) Glycobwlogy 2 , 3 3 7 3 4 3 21. Hara, S., Yamaguchi, M., Takemori, Y., Furuhata, K , Ogura, H., and Nakamura, M. (1989) Anal. Biochem. 179, 162-166 22. Sambrook. J., Fritsch, E.F., and Maniatis, T. (1989) Molecular Cloning: A Laboratory Manual, 2nd ed, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY 23. Shinnick, T.S., Lerner, R. A., and Sutcliffe, J. G. (1981) Nature 293,543-548 24. Olee, T.,Lu,E. W., Huang, D.-F., Soto-Gil, R.W., DeRos, M., Kozin, F.,Carson, D. A,, and Chen, P.P. (1992) J. Exp. Med. 175,8314342 25. Gahm S.J., Fowlkes B. J.,Jameson, S . C., Gascoigne. N. R. J., Cotterman, M. Schwartz, R. H., and Matis, C. A. (1991) Proc. Natl. M., Kanagawa, 0.. Acad. Scz. U. S. A. 88, 10267-10271 26. Jarolim, P., Palek, J., Amato, D., Hassan, K, Sapak, P., Nurse, G. T., Rubin, H. L.. K h a i , S., Sahr, K E., and Liu S . 4 . (1991) Proc.Natl. Acad. Sci. U.S. A. 88, 11022-11026 27. Weichert, M. J., and Chambliss, G. H. (1989) Editorial Comments, U.S. Biochem. Corp. 16,5-6 28. Tabor, S., and Richardson, C.C. (1987) Proc. Natl. A d . Sci. U.S. A. 84, 4767-4771 29. Stanley, P. (1984) Annu. Reu. Genet. 18, 525-552 30. Stanley, P. (1985) Molecular Cell Genetics(Gottesman, M. M.. ed) pp. 745-772, Wiley, New York 31. Bhavanandan. V P., and Katlic, A. W.(1979) J. Bwl. Chem. 264,4000-4008 32. Deutcher, S . L., Nuwayhid, N., Stanley, P., Briles, E. I. B., and Hirschberg, C. B. (1984) Cell 39,29&299 33. Deutscher, S . L., and Hirschberg, C. B. (1986) J. Bwl. Chem. 261,96-100 34. Mapet-Dana, R., Veh, S., Rouche,A., Schauer, R., andMonsigny, M. (1981)Eur: j.Bioehem. 114, 11-16 35. Sandmeyer, S . B., Hansen, L. J., and Chalker, D. L. (1990)Annu.Reu. Genet. 24,491318 36. Goff, S.P. (1987) Methods Enzyrnol. 151,489-502 37. Slightom, J. L., Blechl, A. E., and Smithies, 0.(1980) Cell 21, 627-638 38. Shen, S., Slightom, J. L., and Smithies, 0.(1981) Cell 26, 191-203 39. Kilpatrick, M. W.,Klysik, J., Singleton, C. K, Zarling, D. A,, Jovin, T. M., Hanau, L. H., Erlanger, B. F., and Wells, R. D. (1984) J. Biol. Chem. 259, 7268-7274 40. Hamada, H., and Kakunaga, T.(1982) Nature 298,396-398 41. Hamada, H., Petrino, M. G., and Kakunaga, T.(1982) Proc. Natl. Acad. Sei. U. S.A. 79,6465-6469 42. Hirabayashi, Y., Kasakura, H., Matsumoto, M., Higashi, H., Kato, S.,Kasai, N., and Naiki, M. (1987) Jpn. J. Cancer Res. 78,251-260 43. Schauer, R. (1982) Sialic Acids: Chemistry, Metabolism and Function, Cell Biology Monographs (Schauer, R., ed) Vol. 10, pp. 374386, Springer-Verlag. New York 44. Memck, J. M., Zadarlik, K., and Milgmm, F. (1978) Int. Arch. Allergy Appl. Imrnunol. 57,477-480 45. Fujii, Y., Higashi, H., Ikuta, K , S., and Naiki, M. (1982) Mol. Immunol. IS, 87-94 46. Muchmore, E. A., Milewski, M., Varki, A,, and Diaz, S . (1989) J. Bwl. Chem. 264,20216-20223 47. Shaw, L., and Schauer, R. (1988) Biol. Chem. Hoppe-Seyler 369,417-486 48. Kozutsumi, Y., Kawano, T., Kawasaki, H.,Suzuki, K, Yamakawa, T.,and Suzuki,A. (1991) J. Bioehem. (Tbkyo) 110,429435 49. Kozutsumi, Y., Kawano, T., Yamakawa, T., and Suzuki,A. (1990) J. Biochem. (Zbkyo) 108,704-706 50. Shaw, L., Yousefi, S., Dennis, J. W., and Schauer, R. (1991) Glymconjugate J. 8,434441 51. Hubbard. S . C. (1988) J. Bwl. Chem. 263,1930%19317

Kato.