Proc. Natl. Acad. Sci. USAVol. 89, pp. 329-333, January 1992Biochemistry

Molecular basis of recognition by the glycoprotein hormone-specificN-acetylgalactosamine-transferase

(gonadotropin/pitultary/glycosyltranlsferase/olgo ade/peptide)

PETER L. SMITH AND JACQUES U. BAENZIGER*Department of Pathology, Washington University School of Medicine, St. Louis, MO 63110

Communicated by Oliver H. Lowry, October 7, 1991

ABSTRACT Lutropin (LH) bears asparagine-linked oli-gosaccharides terminating with the unique sequence SO4-4GalNAc31-4GlcNAcfI-2Mana, whereas follitropin (FSH)bears oligosaccharides terminating predominantly with thesequence Siaa-Gall-4GlcNAcI31-2Mana, where Sia is sialicacid. We previously identified a glycoprotein-hormone-specificN-acetylgalacamine-tsferase (GaINAc-transferase) thatrecognizes a peptide-recognition marker(s) present on thecommon glycoprotein hormone a subunit and 13 subunits ofhuman chorionic gonadotropin and LH but not on the (3subunit of FSH. We have now identified an amino acidsequence motif, Pro-Leu-Arg, that is essential for recognitionby the GaINAc-transferase. This tripeptide sequence is found6-9 residues on the amino-terminal side of a glycosylatedasparagine on the a subunit and (3 subunits of LH and humanchorionic gonodatro,pin but is not present on the P subunit ofFSH. The presence of this motif accounts for the differences inLH and FSH oligosaccharide structures. Additional proteinscontaining this recognition motif have been identified and weredetermined to bear sulfated oligosaccharides with the samestructures as those on the glycoprotein hormones, indicatingthat these structures are not restricted to the glycoproteinhormones.

Kupffer cells and that recognizes oligosaccharides terminat-ing with the sequence S04-4GalNAcf91-4GlcNAc/B1-2Mana(17).

In earlier studies, we demonstrated (8) the presence of aglycoprotein-hormone-specific N-acetylgalactosamine-transferase (GalNAc-transferase) in membranes from theanterior pituitary that will transfer GalNAc to the terminalGlcNAc moieties of GlcNAc2Man3GlcNAc2Asn (8). Whenthis oligosaccharide is located on the a subunit or the /3subunits of human (h) CG (hCGf3) or LH (hLH/B), theapparent Km for GalNAc transfer is reduced from 1-2 mM to

330 Biochemistry: Smith and Baenziger

4-ml fraction) and the amount of incorporated [3H]GaiNAcwas quantitated by scintillation counting.

Preparation of Glycopeptides. Sialic acid and galactosewere removed from hCGa, hCGP, and transferrin enzymat-ically (8). hCGa and hCGP were reduced and carboxami-domethylated (RCM) as described (9). hCGa(RCM) (1 mg/mlin 1% NH4CO3) was digested for 18 h at 37C with 5% (wt/wt)sequencing-grade trypsin (Boehringer Mannheim) to obtainglycopeptides hCGa-(52-63) and hCGa-(76-91) and with 5%(wt/wt) sequencing-grade endoprotease Glu-C (BoehringerMannheim) to obtain glycopeptide hCGa-(35-56). hCGJ-(3-20) was obtained by digestion of hCGB3(RCM) with 0.2%sequencing-grade endoprotease Lys-C (Boehringer Mann-heim) and hCGP-(9-20) was obtained by digestion ofhCGj3(RCM) with 5% sequencing-grade trypsin (BoehringerMannheim). The peptide products were identified and iso-lated by reverse-phase HPLC on Synchropak RP-P (50 x 4.6mm, Synchrom) equilibrated in 0.1% trifluoroacetic acid indistilled H20. Initial conditions were maintained for 5 minand followed by a gradient of0-100lo of [0.1% trifluoroaceticacid/56% (vol/vol) acetonitrile/14% (vol/vol) isopropanol/30% (vol/vol) distilled H20] over 50 min. Fractions contain-ing the desired glycopeptides were pooled, Iyophilized, re-suspended in 1 ml of 0.1% trifluoroacetic acid, and chro-matographed as described above except the gradient wasincreased at 1%/min. Peptides were identified and quanti-tated by amino acid analysis. Sequences were confirmed byamino-terminal sequence analysis. The amount of terminalGlcNAc was determined by addition of [3H]galactose usingpurified galactosyltransferase (Sigma) (9).Chemical Modification with 1,2-Cyclohexanedione. Cyclo-

hexanedione-modified hCGa-(35-56) and hCGB-(3-20) wereprepared by dissolving 3 nmol of each glycopeptide in 50 ,ul0.3 M NaOH containing 360 nmol of 1,2-cyclohexanedione(Sigma). After 1 h at 250C in the dark, 30 ,ul of 30% (vol/vol)acetic acid was added followed by 0.5 ml ofdistilled H20. Thecyclohexanedione-modified glycopeptides were eluted assingle peaks, which differed in retention time from theunmodified glycopeptides, when isolated by reverse-phaseHPLC. After modification, arginine was no longer detectedupon amino acid analysis. The "unmodified" glycopeptidesused for comparison were treated identically except that1,2-cyclohexanedione was omitted.

Peptide Synthesis. Peptides were synthesized on a Rampsprocessor (DuPont) according to the manufacturer's instruc-tions. All peptides were purified sequentially by reverse-phase HPLC, S-Sepharose cation-exchange chromatogra-phy, and gel filtration on Sephadex G-10. Peptides werequantitated by amino acid analysis.

RESULTSA 22-Amino Acid Glycopeptide Is Recognized by the Gal-

NAc-Transferase. Our previous studies established that pep-tide recognition is required for specific addition ofGaINAc tooligosaccharides on hCG8, LHB, and the a subunits. Gal-NAc and sulfate are not present on the oligosaccharides ofnative hCG and recombinant LH expressed in CHO cells,due to the absence of the GalNAc- and sulfotransferases inplacental trophoblasts and CHO cells (8, 10). The sialylatedasparagine-linked oligosaccharides on hCG and recombinantLH can be readily converted into substrates for the GalNAc-transferase by removal of terminal sialic acid and galactosemoieties. The lack of dependence on tertiary structure forrecognition by the GaiNAc-transferase indicated that, incontrast to other forms of peptide-dependent recognition, itwould be possible to locate and identify the peptide recog-nition marker by using proteolytic fragments of these hor-mones as acceptors. We therefore prepared proteolytic gly-copeptide fragments from hCGa(RCM) (Fig. 1) andhCGS(RCM) (Fig. 2) for comparison as GaINAc-transferase

substrates. The kinetic parameters determined for hCGa,hCGa(RCM), and hCGa-(35-56), a glycopeptide consistingofamino acids 35-56 ofhCGa including the glycosylation siteat Asn-52, are summarized in Table 1. As we had determined(9), the apparent Km for transfer of GalNAc to oligosaccha-rides on hCGa(RCM) is slightly decreased compared tonative hCGa. The apparent Km of 9.8 ,uM for transfer ofGalNAc to the oligosaccharide acceptor on hCGa-(35-56)does not differ significantly from that for transfer to theoligosaccharides on hCGa(RCM) (Table 1). Therefore, all ofthe information required for recognition of the a subunit bythe GalNAc-transferase resides within this 22-amino acidglycopeptide fragment.

Localizato of the Peptide-Recognition Sequence on hCGa.Additional glycopeptides were prepared to further localizethe amino acids mediating recognition of hCGa by theGalNAc-transferase and to determine if additional recogni-tion sequences might be present. The glycopeptides shown inFig. 1 were compared by determining the time course forGalNAc addition at a substrate concentration of 4 1uM.hCGa(RCM) and hCGa-(35-56), which have nearly identicalcatalytic efficiencies for GalNAc addition (Table 1), incor-porated GaINAc at identical rates (Fig. 1). The glycopeptidefragments hCGa-(52-63) and hCGa-(35-56) contain the iden-tical asparagine-linked oligosaccharide acceptor at Asn-52,yet hCGa-(52-63) did not incorporate detectable levels ofGalNAc until after 100 min of incubation (Fig. 1). We havedetermined (8) that transferrin does not contain a recognitionmarker for the GalNAc-transferase. Detectable levels ofGalNAc incorporation onto transferrin oligosaccharides, likehCGa-(52-63), were also seen only after 100 min of incuba-tion (Fig. 1). The catalytic efficiency for GalNAc addition to

hCG a

1 10 20 30 40 5 60 70 A 90I I I I

hC3.56)1hCGo (52-63)

hCGw (76-91)

o hCXe(RCM)vAhCG(35-56) RAYPTPLRSKKTMLVOKUVTSEAhCGw (5243) NVTSESTCCVAK* hCGe(78-91) VENHTACHCSTCYYHK* trnsftrrhn

16

14'O 120I..1

01

0 25 50 75Minutes

100 125

FIG. 1. Comparison of proteolytic fragments generated fromhCGa(RCM) as substrates for the glycoprotein-hormone-specificGalNAc-transferase. Assays, containing 4 x 10-3 unit of partiallypurified GalNAc-transferase and 240 1M UDPGalNAc, were incu-bated for the indicated time at 37C with hCGa(RCM), hCGa-(35-56), hCGa-(52-63), hCGa-(76-91), or native human transferrin at aconcentration of 4 !M. Amino acid sequences arm shown in single-letter code, with the glycosylated asparagines underlined.

Proc. NatL Acad Sci. USA 89 (1992)

Dow

nloa

ded

by g

uest

on

July

9, 2

021

Proc. Natl. Acad. Sci. USA 89 (1992) 331

hCGP

1 20 40 60 80 100 120 140

AhCGP(3-20) EPLRPRCRPINATLAVEK* hCGp (9-20) CRPINATLAVEKo hCGp (RCM)

16

14

-a 120

&D10

c 8co

_6 6E4

2

0

0 25 50 75 100 125Minutes

FIG. 2. Comparison of proteolytic fragments generated fromhCG(3(RCM) as substrates for the glycoprotein-hormone-specific Gal-NAc-transferase. Assays containing 4 x 10-3 units ofpartially purifiedGalNAc-transferase, 240 ,uM UDPGalNAc, and hCGP(RCM), hCGP-(3-20), or hCGP-(9-20) at a concentration of 4 .tM. The amino acidsequences of hCG3-(3-20) and hCGf-(9-20) are indicated by thesingle-letter code. The glycosylated asparagine at position 13 isunderlined.

oligosaccharides on hCGa(RCM) is more than 100-foldgreater than for addition to the same oligosaccharide struc-tures on either hCGa-(52-63) or transferrin (9). In controlexperiments using bovine j-1,4-galactosyltransferase, an en-zyme that requires only terminal GlcNAc for transfer and isnot influenced by the underlying protein structure (11), equalamounts of galactose were added to the oligosaccharides onhCGa-(35-56) and hCGa-(52-63) confirming that the sameamount of GlcNAc2Man3GlcNAc2Asn was present on bothglycopeptides. The amino acids mediating recognition by theGalNAc-transferase must, therefore, be located betweenamino acids 35 and 51 of the human a subunit.hCGa has a second glycosylation site at Asn-78 (Fig. 1).

The peptide sequence surrounding this site does not resemblethat surrounding Asn-52. A glycopeptide, hCGa-(76-91),containing this glycosylation site was also tested as substratefor the GalNAc-transferase (Fig. 1). GalNAc was transferred

Table 1. Kinetic parameters for addition of GalNAc toasparagine-linked oligosaccharide acceptors on thehormone a subunit

Apparent V.,a CatalyticApparent pmol per h per,g efficiency

Substrate K, ,uM of protein (Vmax/Km)hCGa 14.6 18.26 1.25hCGa(RCM) 8.1 6.06 0.75hCGa-(35-56) 9.8 5.19 0.53

Native hCGa, hCGa(RCM), and hCGa-(35-56) were character-ized as substrates for the glycoprotein-hormone-specific GalNAc-transferase. The apparent K. and Vma. values were determined foreach substrate from secondary plots of Lineweaver-Burk plots atmultiple concentrations of UDPGalNAc.

to hCGa-(76-91) at

332 Biochemistry: Smith and Baenziger

that of unmodified hCGa-(35-56) (Fig. 4A). Likewise mod-ification of the three arginine residues in hCGf-(3-20) re-duced the rate of GaINAc transfer to 10% that of unmodifiedhCGfi-(3-20) (Fig. 4B). In contrast, treatment of eitherhCGa-(35-56) or hCGI3-(3-20) with 1,2-cyclohexanedioneunder conditions that result in a cationic product (13) did notsignificantly alter recognition by the GalNAc-transferase(data not shown). Thus, it appears to be the cationic natureof arginine that is essential for recognition by the GalNAc-transferase. Since the only arginine residues at homologouspositions in hCGa-(35-56) and hCG/3-(3-20) are both a partof the Pro-Leu-Arg motif, the loss of recognition after mod-ification strongly implicates these specific arginine residuesas an essential part of the recognition marker used by theGalNAc-transferase.

Inhibition of the GaINAc-Transferase by Synthetic Peptides.The presence of the peptide recognition marker on glyco-peptide substrates reduces the apparent Km for the oligosac-charide acceptor but is not required for transfer of GalNAc(8), suggesting the peptide marker and the oligosaccharide arerecognized independently. If this is the case, peptides con-taining the recognition marker but no oligosaccharide shouldact as competitive inhibitors of the GalNAc-transferase.Since hCGa-(35-56) contains all ofthe necessary informationrequired for recognition by the GalNAc-transferase, weprepared the identical nonglycosylated peptide synthetically.A peptide with the same sequence as hCGa-(35-56) showeda level of inhibition significantly better than a peptide con-taining the same amino acids in a random order (Fig. 5). Thelevel of inhibition seen with the scrambled peptide wassimilar to that obtained with a number of other unrelatedpeptides of similar size but differing composition. Digestionwith Pronase completely abolished inhibition by either thepeptide with the same sequence as hCGa-(35-56) or thescrambled peptide (data not shown), indicating that in bothcases inhibition is due to peptide and not to nonpeptidecontaminants. Replacement of the Pro-Leu-Arg sequence ofthe peptide representing hCGa-(35-56) with Ala-Ala-Alasignificantly reduced its ability to inhibit the GalNAc-transferase (Fig. 5). Therefore, inhibition ofGalNAc transferby synthetic peptides is sequence-specific, requiring thePro-Leu-Arg motif for maximal inhibition.

DISCUSSIONBased on the evidence presented here, we conclude that thetripeptide sequence Pro-Leu-Arg is the basis for recognition

A RAYPTPLRSKKTMLVQKNVTSE1816

1O4

"-2-_

10

0 25 50 75 100 125minutes

B

0

0

E

0 25 50 75 100 125minutes

FIG. 4. Recognition of hCGa-(35-56) and hCGP-(3-20) by theglycoprotein-hormone-specific GaINAc-transferase is disrupted bychemical modification of arginine residues. Time courses comparingthe addition of GalNAc to oligosaccharides on unmodified (circles)and 1,2-cyclohexanedione-modified (triangles) hCGa-(35-56) (A)and hCGP-(3-20) (B). Assays contained 4 x 10-3 unit of partiallypurified GalNAc-transferase, 240 ,uM UDPGalNAc, and the indi-cated glycopeptide at 4 ,uM. The amino acid sequence of hCGa-(35-56) and hCGP-(3-20) is indicated in the single-letter code. Modifiedarginine residues are indicated (*), and glycosylated asparagines areunderlined.

90

X 600, 50C

coE 40

E3

a

* RAYPTPLRSKKTMLVQKNVTSEA RAYPTAAASKKTMLVQKNVTSEo RRYKVLEPTVLKATQPTKSNMS

0 250 500 750 1000

[MM]FIG. 5. Inhibition of GalNAc transfer to hCGa oligosaccharides

by synthetic peptides. Synthetic peptides corresponding to aminoacids 35-56 ofhCGa (solid circles), amino acids 35-56 ofhCGa withresidues 40-42 (Pro-Leu-Arg) replaced with Ala-Ala-Ala (solid tri-angles), and the same amino acids in a random sequence (opencircles) were compared as inhibitors of GalNAc addition to hCGa.Assays containing 4 x 10-3 unit of partially purified GalNAc-transferase, 100 AM UDPGalNAc, 4,uM hCGa, and the indicatedconcentration of synthetic peptide were incubated at 370C for 90 min.The amino acid sequences are shown in single-letter code.

of hCGa and hCGI3 by the glycoprotein-hormone-specificGalNAc-transferase. This 3-amino acid sequence is located6-9 residues on the amino-terminal side of the asparagineglycosylation sites on hCGf and hCGa, respectively. Acomparison ofthe sequences residing between this tripeptidemotif and the glycosylation site suggests that considerablevariation will be tolerated in this region by the GalNAc-transferase. For most animal species other than humans, thea subunit contains the sequence Pro-Ala-Arg rather thanPro-Leu-Arg (2), suggesting there is also flexibility in theidentity of the middle amino acid of the tripeptide motif;however, the spectrum of acceptable substituents at thisposition remains unclear. Based on homologies with otherGalNAc-transferase substrates (see below), this position maybe restricted to hydrophobic amino acids. Since chemicalmodification of the arginine only results in a loss of recog-nition when its cationic character is destroyed, it is likely thatlysine can be substituted for arginine. We, therefore, proposethat recognition by the GalNAc-transferase requires thepresence of a Pro-Xaa-Arg/Lys motif (where Xaa is mostlikely a hydrophobic residue) located 6-9 residues on theamino-terminal side of an asparagine glycosylation site.

Identification of the tripeptide sequence recognized by theGalNAc-transferase provides a convincing explanation forthe presence of S04-GalNAc on specific glycoprotein hor-mones. In addition to the a subunit, the sequence Pro-Xaa-Arg/Lys is found on LHI3 and CGB but is not present onFSH.8 due to a truncation of its amino terminus relative toLHP and CGB at the gene level. This correlates well with thepresence of sulfated oligosaccharides on the LH dimer and onuncombined a subunits synthesized in the pituitary and theirabsence on the FSH dimer (1, 6, 7). In previous studies wedetermined (9) that combination of FSHP with the a subunitresults in masking of the recognition sequence on the asubunit. The degree to which the Pro-Leu-Arg recognitionsequence on the a subunit is masked most likely accounts forthe extent to which oligosaccharides on FSH dimers bearS044GalNAc. hCGB and hCGa contain the recognitionsequence and the hCG dimer is a substrate for the GalNAc-transferase in vitro; however, hCG synthesized by placental

Proc. Natl. Acad. Sci. USA 89 (1992)

Dow

nloa

ded

by g

uest

on

July

9, 2

021

Proc. Natl. Acad. Sci. USA 89 (1992) 333

trophoblasts does not bear sulfated oligosaccharides becauseneither the glycoprotein-hormone-specific GalNAc-transfer-ase nor the sulfotransferase is expressed in placenta (8).We have searched the NBRF protein data base for addi-

tional glycoproteins containing a Pro-Xaa-Arg/Lys sequencelocated 6-9 residues on the amino-terminal side of an Asn-Xaa-Ser/Thr. Among the glycoproteins identified by thesecriteria, two have been reported to contain sulfated aspar-agine-linked oligosaccharides. Rat proopiomelanocortin andthe glycosylated form of its cleavage product corticotropincontain sulfated asparagine-linked oligosaccharides (14). Atripeptide sequence, Pro-Val-Lys, is located 7 residues on theamino-terminal side of the asparagine glycosylation sitefound within corticotropin. Lipoprotein-associated coagula-tion inhibitor, synthesized by human umbilical vein endothe-lial cells, can be metabolically labeled on its asparagine-linked oligosaccharides with [355]Q4 (G. J. Broze, personalcommunication). Lipoprotein-associated coagulation inhibi-tor has the tripeptide sequence Pro-Phe-Lys located 9 resi-dues on the amino-terminal side of one of its glycosylationsites (15). We have characterized the asparagine-linked oli-gosaccharides on corticotropin synthesized in AtT-20 cells(T. P. Skelton, S. Kumar, P.L.S., and J.U.B., unpublisheddata) and on lipoprotein-associated coagulation inhibitorsynthesized in 293 cells (P.L.S., T. P. Skelton, M. C. Be-ranek, G. J. Broze, and J.U.B., unpublished data) and foundthat >70% of their asparagine-linked oligosaccharides termi-nate with the sequence S04-4GalNAcf31-4GlcNAcp1-2Mana. Thus, the sulfated oligosaccharide structures that weoriginally identified on the pituitary glycoprotein hormonesare also present on other glycoproteins that contain therecognition motif Pro-Xaa-Arg/Lys in close proximity to anasparagine-linked oligosaccharide.

Synthesis of the unique sulfated oligosaccharides found onthe glycoprotein hormones is highly specific and tightlyregulated. The presence of such unique sulfated oligosaccha-rides suggested that they play an important role in thebiologic behavior of glycoprotein such as LH. The sulfatedoligosaccharides found on LH do not affect its bioactivity atthe receptor level but do have a marked impact on circulatoryhalf-life ofLH and, as a result, on its in vivo potency (16). Wehave identified a receptor in hepatic reticuloendothelial cellsthat is specific for oligosaccharides with the terminal se-quence SO4-4GalNAc/31-4GlcNAc.31-2Mana and can ac-count for the rapid clearance of LH and other glycoproteins

bearing these structures (17). The presence of the Pro-Leu-Arg motif and sulfated oligosaccharide structures on otherglycoproteins of diverse function and origin suggests thatthese structures may be of biologic significance for a numberof glycoproteins found in the circulation.

We thank N. L. Baenziger, E. D. Green, and R. Kornfeld forhelpful advice, Dave Donermeyer for amino acid analyses, MarkFrazier of the Washington University Protein Chemistry Lab forpeptide sequencing, and the National Hormone and Pituitary Agencyfor providing purified hCG. This investigation was supported byNational Institute of Diabetes and Digestive and Kidney DiseasesInstitute Grant R01-DK41738. P.L.S. was supported by U.S. PublicHealth Service Grant T32-ES07066 in Mechanisms of Disease inEnvironmental Pathology.

1. Baenziger, J. U. & Green, E. D. (1988) Biochim. Biophys. Acta947, 287-306.

2. Pierce, J. G. & Parsons, T. F. (1981) Annu. Rev. Biochem. 50,465-495.

3. Sairam, M. R. (1983) in Hormonal Proteins and Peptides, ed.Li, C. H. (Academic, New York), Vol. 11, pp. 1-79.

4. Childs, G. V., Ellison, D. G. & Garner, L. L. (1980) Am. J.Anat. 158, 397-409.

5. Childs, G. V. (1984) in Hormonal Control of the Hypothalmo-Pituitary-Gonadal Axis, eds. McKern, K. W. & Noar, Z.(Plenum, New York), pp. 181-198.

6. Green, E. D. & Baenziger, J. U. (1988) J. Biol. Chem. 263,25-35.

7. Green, E. D. & Baenziger, J. U. (1988) J. Biol. Chem. 263,36-44.

8. Smith, P. L. & Baenziger, J. U. (1988) Science 242, 930-933.9. Smith, P. L. & Baenziger, J. U. (1990) Proc. Natl. Acad. Sci.

USA 87, 7275-7279.10. Smith, P. L., Kaetzel, D., Nilson, J. & Baenziger, J. U. (1990)

J. Biol. Chem. 265, 874-881.11. Beyer, T. A., Sadler, J. E., Rearick, J., Paulson, J. C. & Hill,

R. L. (1981) Adv. Enzymol. Relat. Areas Mol. Biol. 52, 23-175.12. Toi, K., Bynum, E., Norris, E. & Itano, H. A. (1967) J. Biol.

Chem. 242, 1036-1043.13. Pathy, L. & Smith, E. L. (1975) J. Biol. Chem. 250, 557-564.14. Bourbonais, Y. & Crine, P. (1985) J. Biol. Chem. 260, 5832-

5837.15. Wun, T. C., Kretzmer, K. K., Girard, T. J., Miletich, J. P. &

Broze, G. J. (1988) J. Biol. Chem. 263, 6001-6004.16. Baenziger, J. U., Kumar, S., Brodbeck, R. M., Smith, P. L. &

Beranek, M. C. (1992) Proc. Natl. Acad. Sci. USA 89,334-338.17. Fiete, D., Srivastava, V., Hindsgaul, 0. & Baenziger, J. U.

(1991) Cell, in press.

Biochemistry: Smith and Baenziger

Dow

nloa

ded

by g

uest

on

July

9, 2

021