SB

ADI

R

isegrgbrtGoeTtogiatwfpYrsii

ir

H

i

0CA

Archives of Biochemistry and BiophysicsVol. 363, No. 1, March 1, pp. 163–170, 1999Article ID abbi.1998.1073, available online at http://www.idealibrary.com on

elf-Glucosylation of Glycogenin, the Initiator of Glycogeniosynthesis, Involves an Inter-subunit Reaction1

my Lin, James Mu, Jie Yang, and Peter J. Roach2

epartment of Biochemistry and Molecular Biology, Indiana University School of Medicine,ndianapolis, Indiana 46202-5122

eceived October 9, 1998, and in revised form December 9, 1998

ogtmcttlopfeeAgp

cgMttvtsaTtb(sbce(tt

Glycogenin is a dimeric self-glucosylating proteinnvolved in the initiation phase of glycogen biosynthe-is. As an enzyme, glycogenin has the unusual prop-rty of transferring glucose residues from UDP-lucose to itself, forming an a-1,4-glycan of around 10esidues attached to Tyr194. Whether this self-lucosylation reaction is inter- or intramolecular haseen debated. We used site-directed mutagenesis ofecombinant rabbit muscle glycogenin-1 to addresshis question. Mutation of highly conserved Lys85 toln generated a glycogenin mutant (K85Q) that hadnly 1–2% of the self-glucosylating activity of wild-typenzyme. Consistent with previous work, mutation ofyr194 to Phe in a GST-fusion protein yielded a mu-

ant, Y194F, that was catalytically active but incapablef self-glucosylation. The Y194F mutant was able tolucosylate the K85Q mutant. However, there was annitial lag in the self-glucosylation reaction that wasbolished by preincubation of the two mutant pro-eins. The interaction between glycogenin subunitsas relatively weak, with a dissociation constant in-

erred from kinetic experiments of around 2 mM. Weropose a model for the glucosylation of K85Q by194F in which mixing of the proteins is followed byate-limiting formation of a species containing bothubunit types. The results provide the most direct ev-dence to date that the self-glucosylation of glycogeninnvolves an inter-subunit reaction. © 1999 Academic Press

Glycogenin is a self-glucosylating protein implicatedn the initiation phase of glycogen biosynthesis [foreviews see (1–3)]. The self-glucosylation generates an

1 This study was supported in part by National Institutes ofealth Grant DK27221.2

tTo whom correspondence should be addressed. E-mail: proach@

upui.edu.

003-9861/99 $30.00opyright © 1999 by Academic Pressll rights of reproduction in any form reserved.

ligosaccharide, composed of glucose residues in a-1,4-lycosidic linkages, that serves as a primer for elonga-ion by glycogen synthase (4). Most work has addressedammalian muscle glycogenin (glycogenin-1) but gly-

ogenin-like genes are quite widely distributed in na-ure (3). For example, Saccharomyces cerevisiae haswo genes, GLG1 and GLG2, that encode self-glucosy-ating proteins necessary for glycogen synthesis in thisrganism (5). In addition, genes or cDNAs predictingroteins with significant sequence similarity have beenound in other organisms, indicating the presence, forxample, of four different genes in Caenorhabditis el-gans, two in Caenorhabditis briggsae, and four inrabidopsis thaliana. Recently, a second mammalianene, glycogenin-2, was described that is expressedredominantly in liver, heart, and pancreas (6).The glycogenins that have been studied biochemi-

ally share the property of catalyzing the transfer oflucose from UDP-glucose, in a reaction requiringn21, to form an oligosaccharide chain from around 8

o 20 residues. Glycogenin is an unusual enzyme inhat the product of the reaction is a covalently modifiedersion of itself. However, unlike certain types of au-ocatalytic processes, there is turnover at the activeite of the glycogenin, since multiple glucose transfersre required for the formation of the reaction product.he initial attachment of glucose is via a glucose-1-O-yrosine linkage, rare among glycoproteins, and in rab-it muscle glycogenin the modified residue is Tyr1947, 8). Both the initial modification of Tyr194 and theubsequent transfer of glucose residues are believed toe mediated by glycogenin (9). Glycogenin can alsoatalyze the transfer of glucose residues to small mol-cule acceptors, such as maltose or maltose derivatives10–13). Mutation of Tyr194 eliminates the ability ofhe glycogenin to self-glucosylate but does not disableransglucosylation of other acceptors (14, 15). If any-

hing, transglucosylation is enhanced, possibly due to

163

lg

atacocglrsal(ecqn

ssipnsrptgec

E

C

m(epcCcpTdwmdtedap

gH(rfcvSa

DdGGomHe

P

p

Faleicw

164 LIN ET AL.

oss of the competing and sterically favored self-lucosylation (12).There has been some debate over the kinetic mech-

nism of the self-glucosylation, as to whether the reac-ion is intra- or intermolecular (see Fig. 1). Pitcher etl. (16) reported that the specific activity for self-glu-osylation by glycogenin was essentially independentf glycogenin concentration, a result that was dupli-ated later by Cao et al. (14) using recombinant glyco-enin. These results would imply that the self-glucosy-ation reaction is intramolecular. Alonso et al. (17)eported that at lower concentrations of glycogenin, thepecific activity did become concentration dependentnd proposed that the reaction could be intermolecu-ar. Gel filtration suggested that glycogenin is a dimer18), a result also confirmed in the present study. Caot al. (12) analyzed the X-ray diffraction of small gly-ogenin crystals and, although the data were inade-uate for solution of the three-dimensional structure,

IG. 1. Possible mechanisms for glycogenin self-glucosylation. Fordimeric glycogenin molecule, glucosylation can be either intramo-

ecular (A) or intermolecular (B). An intramolecular reaction could beither inter-subunit (i) or intra-subunit (ii). The glycogenin subunits depicted as a circle, with the site of glucosylation, Tyr194, indi-ated by “Y.” The active site is labeled with “K,” symbolizing Lys85hich is important for catalytic activity, as discussed in the text.

oncrystallographic symmetries were detected thatpc

uggested the presence of dimers. An intramolecularelf-glucosylation reaction could, therefore, be eitherntra- or inter-subunit (Fig. 1). The purpose of theresent study was to investigate further the mecha-ism of self-glucosylation. Our conclusion is that theelf-glucosylation of glycogenin involves inter-subuniteactions, most likely occurring within a dimeric com-lex as an intramolecular reaction under most condi-ions. We also propose that the interaction betweenlycogenin subunits is relatively weak, allowing forxchange of subunits between different oligomericomplexes.

XPERIMENTAL PROCEDURES

onstruction of Expression VectorsH6GN-K85Q. A segment of the coding region of rabbit skeletaluscle glycogenin was amplified by the polymerase chain reaction

PCR) (94°C, 1 min; 55°C, 2 min; 72°C, 3 min for 25 cycles) using anxisting wild-type construct, pET15b/GN (12), as template. Thislasmid expresses a His-tagged version of glycogenin in Escherichiaoli. The sense primer was 59-AAGAGGCCTGAGTTGGGTGTCA-ACTGACCCAACTG-39. A StuI site is underlined and the italicizedodon causes a mutation of residue 85 from Lys to Gln. The antisenserimer was 59-CAGCCGGATCCTACTGGAGGTAA GTGTCAAGTT-CTTC-39, where a BamHI site is underlined. The PCR product wasirectly ligated into the PCR vector (Invitrogen). Competent cellsere transformed with the ligation mixture and selected by kana-ycin (Sigma). The sequence of the PCR product was determined by

ideoxy sequencing to ensure the fidelity of DNA amplification andhe presence of the mutations. Digestion with StuI and BamHIxcised a segment of glycogenin coding sequence, containing theesired mutation, that was inserted back into the corresponding StuInd BamHI sites of the wild-type pET15b/GN vector via a three-iece ligation to generate pET15b/GNK85Q.Calmodulin binding peptide (CBP)3-GN. The coding region for

lycogenin was excised from pET15b/GN by digestion with NdeI andindIII and ligated into the XhoI/HindIII site of the pCAL-N vector

Stratagene), which encodes a fusion protein with a CBP tag. Theesulting construct was designated pCAL-N/GN. A BglII–HindIIIragment containing the CBP tagged coding sequence was then ex-ised and cloned into BglII and HindIII sites in the pACYC184-1ector, obtained from Anna DePaoli-Roach, Indiana Universitychool of Medicine. This vector has a p15A origin of replication andchloramphenicol resistance marker.GST-GN and GST-GN-Y194. Plasmid pGEX-HIBADH was fromr. Robert A. Harris, Indiana University School of Medicine. It waserived from pGEX-KG (19) and contains an NdeI site in frame withST after the BamHI site and a glycine-rich linker (LVPRRI-GGGH). Wild-type and tyrosine-mutated glycogenin-1 were cut outf wild-type pET-15b-GN or the corresponding mutant with Tyr194utated to Phe (12) with NdeI–HindIII and ligated into a pGEX-IBADH vector fragment generated by digestion with the same

nzymes.

urification of Recombinant ProteinsHis-tagged proteins. BL21/DE3 cells were transformed with the

lasmid carrying the glycogenin expression vector and grown in

3 Abbreviations used: CBP, calmodulin binding peptide; PMSF,

henylmethylsulfonyl fluoride; DTT, dithiothreitol; CD, circular di-hroism.

M0ifbTnt(Pwtwipdowwo

wsmbmb

G(pbr(mpcteGma

[pHctswsa60d

zfe

wtcURpt

T

tbPtbJcdtmbt

ms(T

R

staphscwts

Fceo

165INTER-SUBUNIT GLUCOSYLATION OF GLYCOGENIN

9ZB medium. When the A600 of an 800-ml culture reached 0.7–0.8,.4 mM IPTG was added and cells were grown at 30°C for 3 h tonduce expression. Cells were harvested by centrifugation at 4500gor 10 min. The cell pellet was resuspended with homogenizinguffer and passed through a French pressure cell at 1000 lb/in2 twice.he crude extract was centrifuged at 19,000g for 20 min. The super-atant was collected and the crude extract was absorbed in a batcho 2 ml Ni–NTA–agarose beads (Qiagen) equilibrated with Buffer A50 mM Hepes, pH 7.5, 2 mM benzamidine, 0.1 mM TLCK, 1 mMMSF) at 4°C for 1 h. The beads were then loaded into a column,ashed successively with 10 ml Buffer A plus 5 mM imidazole and

hen 10 ml Buffer A plus 10 mM imidazole. The protein was elutedith 5 ml Buffer A plus 50 mM imidazole followed by 5 ml 100 mM

midazole in Buffer A. The fractions containing glycogenin wereooled and dialyzed against 50 mM Hepes, pH 7.5, 2 mM benzami-ine, 1 mM EDTA, 1 mM DTT. The dialyzed sample was then loadednto a Q Sepharose Fast Flow column (5 ml, Pharmacia). Afterashing with 50 ml 50 mM NaCl in Buffer A, glycogenin was elutedith a linear gradient of 50–500 mM NaCl in Buffer A at a flow ratef 0.3 ml/min.CBP-tagged protein. Cell extract, prepared as described above,as applied to calmodulin–agarose resin (Strategene) by batch ab-

orption. After washing with 50 mM Hepes, pH 7.5, 150 mM NaCl, 1M DTT, 2 mM CaCl2, 1 mM PMSF, 0.1 mM TLCK, and 2 mM

enzamidine, the column was eluted with 50 mM Hepes, pH 7.5, 1M DTT, 2 mM EGTA, 1 mM PMSF, 0.1 mM TLCK, and 2 mM

enzamidine.GST-tagged proteins. Expression and purification of GST orST-fusion proteins were performed according to Guan and Dixon

19). Basically, 400 ml bacterial cells expressing GST alone or fusionroteins were cultured at 37°C and induced with 0.1 mM isopropyl-D-thiogalactopyranoside when absorption of 0.6 at 600 nm waseached. Harvested cell pellets were resuspended in GST lysis buffer140 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, 1

M DTT, 2 mM benzamidine, and 0.1% Tween 20) and broken byassing through a French pressure cell followed by a 20-min 10,000gentrifugation. The resulting supernatant was loaded onto a gluta-hione–agarose (Sigma) column. Columns (;1 ml) were then washedxtensively with GST lysis buffer before proteins were eluted withST elution buffer (50 mM Tris–HCl, pH 8.0, 5 mM glutathione, 2M benzamidine, and 1 mM DTT). Eluted samples were dialyzed

gainst GST lysis buffer to remove glutathione.Glucosylation assays. The assay measures the amount of

14C]glucose incorporated into glycogenin essentially as describedreviously (14). The reaction was typically carried out in 50 mMepes, pH 7.5, 2 mM DTT, 5 mM MnCl2, and 20 mM UDP-[14C]glu-

ose. Two methods were applied to quantitate glycogenin glucosyla-ion. In one, aliquots of the reaction mixture were spotted onto aquare of P81 chromatography paper which was washed three timesith 0.5% phosphoric acid for 20 min each. The paper was dried and

ubjected to scintillation counting. Alternatively, an aliquot wasdded to one-quarter volume of 53 loading buffer [60 mM Tris, pH.8, 10% (v/v) glycerol, 0.01% bromophenol blue, 2% (w/v) SDS, and.7 M b-mercaptoethanol] and subjected to SDS–PAGE and autora-iography.

IG. 2. Scheme for the purification of glycogenin dimers containinoexpression of H6GN-K85Q and CBP-GN in E. coli would yield a m

xpression. Purification over calmodulin–agarose should yield only dimeught then to yield only heterodimers.

Analysis of enzyme kinetic data. For a reaction catalyzed by en-yme at concentration ET with substrate at concentration ST, theollowing equation holds regardless of the relative concentrations ofnzyme and substrate:

Km 5~Vmax 2 v!

vST 2

~Vmax 2 v!

VmaxET ,

here Km is the Michaelis constant, v is the reaction rate, and Vmax ishe rate corresponding to saturation of the enzyme at high substrateoncentration. This treatment is that of Dr. Arthur Schulz, Indiananiversity School of Medicine, based on the earlier formulation ofeiner (20). The value of Vmax is estimated from a double-reciprocallot, which is concave upward, as the reciprocal of the intercept withhe ordinate. Then, for v5 Vmax/2,

ST 5 Km 2 ET / 2.

hus, Km can be interpolated from a plot of v versus ST.Circular dichroism (CD). Wild-type H6GN and H6GN-K85Q mu-

ant glycogenin were dialyzed against 2 liters of 10 mM phosphateuffer, pH 7.5, for 4 h and then, after a buffer change, overnight.rotein concentrations of the dialyzed samples were measured andhe proteins were diluted to 0.5 or 0.25 mg/ml in 10 mM phosphateuffer, pH 7.5. CD spectra from 190 to 250 nm were recorded on aasco 720 spectropolarometer at room temperature. A 0.05-cm opti-al path cuvette was used for all measurements. A blank spectrum ofialysate buffer was subtracted from the corresponding protein spec-rum. The results were normalized and expressed as mean residueolar ellipticity. The estimation of the secondary structure was

ased on the reference CD spectrum method of Yang et al. (21) usinghe program software SSE-338 provided by Jasco Corp.

Other materials and methods. Protein concentration was deter-ined by the method of Bradford (22) using bovine serum albumin as

tandard. SDS–PAGE followed essentially the method of Laemmli23). UDP-[14C]glucose was prepared essentially by the method ofan (24).

ESULTS

Experimental strategy. The initial strategy for thistudy was (i) to generate a mutant form of glycogeninhat was catalytically inactive, (ii) to coexpress activend inactive glycogenin subunits in E. coli, and (iii) tourify wild-type/mutant heterodimers (Fig. 2). In sucheterodimers, glucosylation of the mutant proteinhould only be possible if inter-subunit transfer of glu-ose occurred. Alternatively, glucosylation only of theild-type subunit would imply an intra-subunit reac-

ion. The wild-type and mutant proteins would be con-tructed with different protein tags that would make

ctive and inactive subunits. If glycogenin dimers were stable, thenure of homo- and heterodimers, depending on the relative levels of

g aixt

rs containing CBP-GN. Subsequent purification on Ni–NTA–agarose

trcp

gWddm4wpe

it

lieTlripTnTa4t(imsawtplvsc(ca3

twlCo

Frsaaccr

FTf(s

Fdqw1H

166 LIN ET AL.

he polypeptides differ in size and would also aid pu-ification. Successive passage over two affinity matri-es, each specific for one of the peptide tags, shouldurify only heterodimers.Characterization of recombinant glycogenins and

eneration of the catalytically impaired K85Q mutant.ild-type glycogenin with an NH2-terminal hexahisti-

ine tag, H6-GN, was produced in E. coli essentially asescribed previously (12) and was purified close to ho-ogeneity (Fig. 3). The protein self-glucosylated (Fig.

) and had a specific activity of 2–4 nmol/min/mg,hich is in the range observed previously (12, 14). Thisrotein serves as a reference or control in a number ofxperiments.In the absence of any protein chemical or structural

nformation about residues involved in glycogenin ca-alysis, the alignment of glycogenin and glycogenin-

IG. 3. Analysis of wild-type and mutant glycogenins. Purifiedecombinant glycogenins were analyzed by SDS–PAGE followed bytaining with Coomassie blue. Lane 1, H6GN (2 mg) after Ni–NTA–garose chromatography; lane 2, H6GN-K85Q (2 mg) after Ni–NAT–garose chromatography; lane 3, H6GN (6.5 mg) after Q Sepharosehromatography; lane 4, H6GN-K85Q (6.5 mg) after Q Sepharosehromatography; lane 5, GST-GN-Y194F (6.5 mg) after chromatog-aphy on glutathione–agarose.

IG. 4. Self-glucosylation by wild-type and mutant glycogenins.he indicated protein was incubated in a self-glucosylation reaction

or 5 min as described under Experimental Procedures and aliquots

N0.33 mg protein) were subjected to SDS–PAGE. An autoradiogram ishown. Lanes 1 and 2, H6GN; lanes 3 and 4, H6GN-K85Q.

ike sequences from a variety of organisms was exam-ned [see (3)]. A Lys residue was found to be conserved,ven in the most phylogenetically distant sequences.his Lys residue, Lys85 in rabbit muscle glycogenin, is

ocated in the catalytically active NH2-terminal 200esidues of the protein. Lys was mutated to Gln, tontroduce a side chain of similar size but without theositive charge, in a His-tagged protein, H6GN-K85Q.he recombinant protein was purified close to homoge-eity by similar means as the wild-type H6-GN (Fig. 3).he resulting protein had very little activity whenssayed in the standard self-glucosylation assay (Fig.). More detailed analyses indicated that the activity ofhe mutant was 1–2% of that of wild-type glycogeninsee later in Fig. 8). The mutant was also severelympaired in its ability to transfer glucose to the small

olecule acceptor n-dodecyl b-D-maltoside (data nothown), suggesting that any loss of activity was due tocatalytic defect rather than an inability to interactith an acceptor protein. When analyzed by gel filtra-

ion on Superose 12, both H6-GN and H6GN-K85Qroteins behaved similarly, running with apparent mo-ecular weight of 80,000–100,000, consistent with pre-ious reports that the protein is a dimer (data nothown). Analysis of wild-type and mutant proteins byircular dichroism resulted in almost identical spectradata not shown), indicating that the mutation had notaused any gross structural derangement. Using thenalysis of Yang et al. (21), the data would predict3–37% a helix and 23–24% b sheet.Wild-type glycogenin was also fused NH2-terminally

o a CBP by expression from a pACYC184-1 vectorhich bears a p15A replication origin. This vector gave

ower levels of expression in E. coli and the protein,BP-GN, purified by Ca21-dependent binding to calm-dulin–agarose, was less pure than H -GN (Fig. 5).

IG. 5. Analysis of CBP-glycogenin. Self-glucosylation assays, asescribed under Experimental Procedures, were run for 5 min. Ali-uots (2 mg of glycogenin) were analyzed by SDS–PAGE and stainedith Coomassie blue (A) and subjected to autoradiography (B). Lane, H6GN; lane 2, CBP-GN; lane 3, H6GN-K85Q; lane 4, CBP-GN plus

6GN-K85Q.

6

otably, there was a significant contaminant of ;60

kltP

HaptrqNwi(olsrttlsnhhCAHtcrnte

ochp

WHlHrgmaStasrwotgwlsolkt

FHtiH

FcfHrw1l

167INTER-SUBUNIT GLUCOSYLATION OF GLYCOGENIN

Da. Nonetheless, CBP-GN was active for self-glucosy-ation and the glucosylated species was slightly largerhan H6-GN and was easily distinguished on SDS–AGE (Fig. 5).Coexpression of H6GN-K85Q and CBP-GN in E. coli.aving engineered differentially tagged active and in-ctive glycogenin polypeptides, we coexpressed the tworoteins in E. coli by taking advantage of the fact thathe corresponding plasmids had different origins ofeplication. The expressed protein was subjected se-uentially to calmodulin–agarose chromatography andi–agarose chromatography. The resulting productas analyzed by self-glucosylation and, contrary to

nitial expectations, both subunits were clearly labeledFig. 6). If we had purified stable heterodimers, onlyne or the other of the subunits should have beenabeled if the reaction was intramolecular. There areeveral possibilities to explain the result. First, theeaction could have been an intermolecular one be-ween dimers. The kinetics, however, suggest an in-ramolecular reaction under most conditions (also, seeater). Second, within a dimer, both inter- and intra-ubunit reactions could have operated. This mecha-ism seems less likely. Third, we may not have purifiedeterodimers as intended. We favor the last-mentionedypothesis because the relative stoichiometry ofBP-GN and H6GN-K85Q was not 1:1 as expected.fter the calmodulin–agarose column, the CBP-GN:6GN-K85Q ratio was more like 3:1 in the eluted pro-

ein (not shown). After the subsequent Ni–agarosehromatography, the proportions of the proteins wereeversed, so that the ratio CBP-GN:H6GN-K85Q wasow 1:2 to 3. Our interpretation of these data is thathe interactions between glycogenin subunits are weaknough that subunit exchange occurs on the time scale

IG. 6. Self-glucosylation of CBP-GN/H6GN-K85Q. CBP-GN and6GN-K85Q were coexpressed in E. coli and purified as described in

he text and then subjected to self-glucosylation. An autoradiogram

Hs shown. Lane 1, CBP-GN/H6GN-K85Q (3 mg); lane 2, purified

6GN-K85Q (1 mg); lane 3, CBP-GN (2 mg).

f the purification. Thus, any heterodimers isolatedould reequilibrate to reform mixtures of hetero- andomodimers. Further support for this hypothesis isresented below.Intersubunit glucosylation reactions in glycogenin.e also found that simple mixing of CBP-GN and6GN-K85Q in a self-glucosylation assay resulted in

abeling of both species (Fig. 5). While labeling of6GN-K85Q can only be explained by an inter-subunit

eaction, we can make no conclusions regarding thelucosylation of CBP-GN. Therefore, we constructed autant glycogenin in which Tyr194 is altered to Phe

nd the glycogenin sequence is fused to glutathione-transferase (GST-GN-Y194F). The recombinant pro-ein was purified by chromatography on glutathione–garose (Fig. 3). The preparation contained two majorpecies, GST-GN-Y194F and a smaller polypeptide cor-esponding to glutathione S-transferase. Consistentith previous work (14, 15), the protein was incapablef self-glucosylation (Fig. 7). The corresponding wild-ype GST-fusion protein, GST-GN, was active for self-lucosylation (Fig. 7). When H6GN-K85Q was mixedith either GST-GN or GST-GN-Y194F, we observed

abeling of H6GN-K85Q. With GST-GN-Y194F, the re-ult can only be explained by an inter-subunit transferf glucose. We therefore used this system, the glucosy-ation of H6GN-K85Q by GST-GN-Y194F, to study theinetics of the reaction further. First, we noted thathere was a lag in the time course of glucosylation of

IG. 7. Glucosylation of H6GN-K85Q by GST-GN-Y194F. Self-glu-osylation assays were as described under Experimental Proceduresor 16 min except that UDP-glucose was 77 mM. The amounts ofis-tagged or GST-tagged proteins run per lane were 3.2 and 9.4 mg,

espectively. The proteins were analyzed by SDS–PAGE and stainedith Coomassie blue (A) and subjected to autoradiography (B). Lane, GST-GN plus H6GN-K85Q; lane 2, GST-GN; lane 3, H6GN-K85Q;ane 4, GST-GN-Y194F plus H6GN-K85Q; lane 5, GST-GN-Y194F.

6GN-K85Q that was not present when wild-type

HKilicawmt

HcatrHtrcttltsrc

D

vgiodcNawgtt

FYgqt(Kcms

FtKmfirpac

FtKaHttpsci

168 LIN ET AL.

6GN was analyzed (Fig. 8). Preincubation of H6GN-85Q with GST-GN-Y194F eliminated the lag suggest-

ng the formation of an intermediate that was rateimiting for glucosylation (Fig. 8). In subsequent exper-ments, a preincubation was included. When the glu-osyl acceptor H6GN-K85Q concentration was variedt constant GST-GN-Y194F, the glucosylation reactionas saturable at higher concentrations (Fig. 9). Esti-ates of the Vmax were made from the intersection with

he ordinate in double-reciprocal plots. Considering

IG. 8. Time course of H6GN-K85Q glucosylation by GST-GN-194F. Glycogenins were incubated under standard conditions forlucosylation as described under Experimental Procedures and ali-uots were analyzed for glucosylation at the indicated times usinghe filter assay. Wild-type H6GN (filled triangles) and H6GN-K85Qopen triangles) were present at 2 and 7.1 mM, respectively. H6GN-85Q (7.1 mM) was mixed with GST-GN-Y194F (7.1 mM) and glu-

osylation followed with (filled circles) or without (open circles) a 5in preincubation. The inset shows only the first 10 min to empha-

ize the lag.

IG. 9. Dependence of glucosylation rate on H6GN-K85Q concen-ration. The glucosylation of the indicated concentration of H6GN-85Q by 5 mM (triangles) or 2.5 mM (circles) GST-GN-Y194F waseasured as described under Experimental Procedures using thelter assay with aliquots of 8 ml spotted onto the P81 papers. Double-eciprocal plots were extrapolated to Vmax values of 2.8 and 4.5mol/min/assay for 2.5 and 5 mM GST-GN-Y194F, respectively. The

dssay was for 20 min at 30°C, and glucosylation is referred to theoncentration of active subunits.

6GN-K85Q as an enzyme substrate present at a con-entration comparable to that of the enzyme, a Km ofpproximately 2 mM can be calculated (see Experimen-al Procedures). The concentration dependence of theeaction was also examined by simultaneously varying

6GN-K85Q and GST-GN-Y194F at equimolar concen-rations (Fig. 10). Expressed per active subunit, theate of glucosylation was essentially invariant over theoncentration range examined and was about 25% ofhe wild-type level. This result is suggestive of an in-ramolecular as opposed to an intermolecular glucosy-ation reaction in this concentration range. With wild-ype glycogenin, we observed relative invariance of theelf-glucosylation rate over much of the concentrationange studied, although there was possibly some de-rease at the lowest concentrations tested (Fig. 10).

ISCUSSION

The first significant result of this work is the obser-ation that mutation of Lys85 of rabbit muscle glyco-enin to Gln results in a protein which was severelympaired regarding self-glucosylation, with only 1–2%f the wild-type activity under standard reaction con-itions. Transglucosylation of the small molecule ac-eptor n-dodecyl b-D-maltoside was similarly defective.onetheless, the protein was stable, could be purified,nd, by gel filtration, behaved as a dimer just likeild-type glycogenin. Analysis by circular dichroismave a spectrum for the mutant virtually identical tohat of the wild-type protein. We therefore concludehat the impaired enzymatic activity is not due to a

IG. 10. Concentration dependence of glycogenin self-glucosyla-ion. Wild-type H6GN (circles) or an equimolar mixture of H6GN-85Q and GST-GN-Y194F (squares) was analyzed for glucosylations described under Experimental Procedures using the filter assay.

6GN-K85Q and GST-GN-Y194F were preincubated for 5 min prioro initiating the reaction. Assays were for 5 min at room tempera-ure. Glucosylation is referred to the concentration of active subunitsresent. Glycogenin concentration on the abscissa is referred to totalubunits. The dashed line shows the calculated amount of dimerorresponding to the monomer concentration of the abscissa assum-ng a dissociation constant of 2 mM.

isruption of the overall structural integrity of the

pismsqcLdstorStlpfrcMvitIiaptbacafi

ttoiaTmswaltsttmivtp(

wWuiiv

ltptpaitcuiwescdpcwsopfweh

slicttiacopiiccptc

169INTER-SUBUNIT GLUCOSYLATION OF GLYCOGENIN

rotein. The choice of Lys85 was based on the fact thatt was the only basic residue that was absolutely con-erved, even in the glycogenin-like sequences most re-ote from mammalian glycogenin, in a comparison of

uch sequences by Roach and Skurat (3). Subse-uently, several other sequences have appeared, allontaining a Lys in correspondence with rabbit muscleys85. This highly conserved, positively charged resi-ue is a candidate to interact with the phosphate in theubstrate UDP-glucose and so its mutation might in-erfere with substrate binding. Lys85 is located in onef the most highly conserved regions of glycogenin andelated proteins, domain II according to Roach andkurat (3). This residue is only 16 residues NH2-erminal of a motif, -D-X-D-, found in all glycogenin-ike sequences. This same motif has recently been re-orted to be common to a number of glycosyl trans-erases and to be essential for activity (25, 26). Theseesidues were implicated in Mn21 binding (25, 26),onsistent with the requirement of glycogenin forn21. It is therefore quite possible that Lys85 is in-

olved in the active site of glycogenin but understand-ng its exact function will require the determination ofhe three-dimensional structure of the protein. DomainI of glycogenin and related proteins also contains anmperfect Leu zipper, as was pointed out by Cheng etl. (5). Thus, one might have hypothesized a role inrotein–protein interactions, such as in the dimeriza-ion of glycogenin subunits. Whether such a role woulde compatible with being close to the active site alsowaits more detailed structural information about gly-ogenin. We do note that mutation at Lys85 did notffect the dimerization of glycogenin as judged by gelltration.Perhaps the most important experimental finding of

his study is that the catalytically crippled K85Q mu-ant could be glucosylated by the Y194F mutant formf the protein. First, this is another indication of thentegrity of the K85Q mutant. It has been shown that

synthetic peptide with the sequence surroundingyr194 cannot be glucosylated by wild-type or Y194Futant glycogenin (14). Jansson et al. (27) obtained a

imilar result and additionally tested similar peptidesith glucose or maltose residues attached to the equiv-lent of Tyr194. These glycopeptides could be glucosy-ated by glycogenin but only to very low stoichiome-ries. Therefore, it is likely that a higher degree oftructure in a protein acceptor is needed for optimalransfer. Second, and more importantly, the glucosyla-ion of K85Q by Y194F sheds light on the molecularechanism of the self-glucosylation of glycogenin. It

mplies that the self-glucosylation of glycogenin in-olves an inter-subunit reaction. If the reaction is in-ramolecular, as we infer from the concentration inde-endence over quite a wide range of concentrations

Fig. 10), then this intra-subunit reaction would be

ws

ithin a given glycogenin dimer [Model A (ii) in Fig. 1].e had previously proposed this model for intramolec-

lar, inter-subunit self-glucosylation, as opposed to anntra-subunit one, based only on esthetic and teleolog-cal considerations (12). The present work now pro-ides experimental support.Using the K85Q mutant as the acceptor for glucosy-

ation by a Y194F mutant, we observed that the reac-ion had a lag that was absent with the wild-typerotein. The lag disappeared upon preincubation of thewo mutant proteins implying that, with the mutantroteins, there was a step before the glucosylation re-ction that was rate limiting. Since glucosyl transfer isnter-subunit in this system, a prime possibility is thathe rate-limiting step is the formation of oligomersontaining both mutant proteins. With the systemsed, there is a practical complication because GST

tself can dimerize (28). Therefore, we do not knowhether the active species is a K85Q/Y194Y het-

rodimer or some higher order complex containing bothubunit types. Probably, however, subunits can ex-hange, consistent with the idea that glycogeninimers are not very stable. This suggestion is sup-orted by the kinetic data derived from varying theoncentration of the K85Q substrate protein, fromhich an apparent Km of ;2 mM was calculated. Weak

ubunit interactions can also explain why purificationf heterodimers by successive affinity chromatogra-hies, specific for each subunit of the dimer, selectedor the binding subunit. A population of heterodimersould equilibrate into a mixture of homo- and het-rodimers, making it impossible ever to isolate pureeterodimers, as had been hoped.As mentioned in the Introduction, there has been

ome debate over the mechanism for the self-glucosy-ation of glycogenin, specifically as to whether it isnter- or intramolecular. Alonso et al. (17) reported aoncentration dependence of glycogenin self-glucosyla-ion at very low protein concentrations. They posedhis result as challenging whether the reaction wasntramolecular or not and considered the matter to ben open question. We believe that there is little dis-repancy between the results of Alonso et al. (17) andur current model for glycogenin self-glucosylation, es-ecially taking into account that the subunit–subunitnteractions are not especially strong. Any dimer thats not infinitely stable must dissociate at low enoughoncentrations and so any activity– concentrationurve that depends on the presence of a dimer mustass through the origin4 (see Fig. 10). At concentra-ions below the dissociation constant, there would be aoncentration dependence to the formation of the ac-

4 The exception would be if the reaction were intra-subunit in

hich case the reaction would be independent of oligomerization

tate.

tTlurssrgagist

bnpfgmmaogRdg

A

DAe

R

1

1

1

1

1

1

1

1

1

1

2

2

2222

2

2

170 LIN ET AL.

ive dimer and hence to the self-glucosylation rate.his is equivalent to saying that there is an intermo-

ecular, bimolecular reaction between isolated sub-nits at these low concentrations. In our hands, as weeduce glycogenin concentration, we reach the limits ofensitivity for our detection system before we clearlyee a statistically significant decrease in glucosylationate (Fig. 10). Alonso et al. (17), working with humanlycogenin, succeeded in monitoring a decrease in ratet very low glycogenin concentrations. A model forlycogenin function with an active dimer capable ofnter-subunit glycosylation and with relatively weakubunit–subunit interactions accommodates most ofhe results.

It is possible that the weakness of the interactionsetween glycogenin monomers has a physiological sig-ificance. After self-glucosylation has generated arimed glycogenin subunit, there is no longer any needor the existence of glycogenin dimers. The action oflycogen synthase and branching enzyme will lead toature glycogen molecules attached to each glycogeninonomer. Perhaps, the weakness of the interactions

llows for mature glycogen molecules to separate. Theligomeric state of glycogenin when attached to glyco-en is not known and might be difficult to determine.egardless, the results of this study provide the mostirect evidence to date that self-glucosylation by glyco-enin involves an inter-subunit reaction.

CKNOWLEDGMENTS

We thank Drs. Anna DePaoli-Roach and Thomas Hurley of thisepartment for many helpful discussions. We are grateful to Dr.rthur Schultz, also of this Department, for help formulating thenzyme kinetic analysis.

EFERENCES

1. Smythe, C., and Cohen, P. (1991) Eur. J. Biochem. 200, 625–631.2. Alonso, M. D., Lomako, J., Lomako, W. M., and Whelan, W. J.

(1995) FASEB J. 9, 1126–1137.3. Roach, P. J., and Skurat, A. V. (1997) Prog. Nucleic Acids Res.

Mol. Biol. 57, 289–316.4. Skurat, A. V., and Roach, P. J. (1995) in Diabetes Mellitus: A

Fundamental and Clinical Text (LeRoith, D., Olefsky, J. E., andTaylor, S., Eds.), pp. 213–222, Lippincott, Philadelphia.

2

5. Cheng, C., Mu, J., Farkas, I., Huang, D., Goebl, M. G., andRoach, P. J. (1995) Mol. Cell. Biol. 15, 6632–6640.

6. Mu, J., Skurat, A. V., and Roach, P. J. (1997) J. Biol. Chem. 272,27589–27597.

7. Rodriguez, I. R., and Whelan, W. J. (1985) Biochem. Biophys.Res. Commun. 132, 829–836.

8. Smythe, C., Caudwell, F. B., Ferguson, M., and Cohen, P. (1988)EMBO J. 7, 2681–2686.

9. Alonso, M. D., Lomako, J., Lomako, W. M., Whelan, W. J., andPreiss, J. (1994) FEBS Lett. 352, 222–226.

0. Lomako, J., Lomako, W. M., and Whelan, W. J. (1990) FEBSLett. 264, 13–16.

1. Manzella, S. M., Roden, L., and Meezan, E. (1994) Anal. Bio-chem. 216, 383–391.

2. Cao, Y., Steinrauf, L. K., and Roach, P. J. (1995) Arch. Biochem.Biophys. 319, 293–298.

3. Manzella, S. M., Roden, L., and Meezan, E. (1995) Glycobiology5, 263–271.

4. Cao, Y., Mahrenholz, A. M., DePaoli-Roach, A. A., and Roach,P. J. (1993) J. Biol. Chem. 268, 14687–14693.

5. Alonso, M. D., Lomako, J., Lomako, W. M., and Whelan, W. J.(1994) FEBS Lett. 342, 38–42.

6. Pitcher, J., Smythe, C., and Cohen, P. (1988) Eur. J. Biochem.176, 391–395.

7. Alonso, M. D., Lomako, J., Lomako, W. M., and Whelan, W. J.(1995) J. Biol. Chem. 270, 15315–15319.

8. Smythe, C., Watt, P., and Cohen, P. (1990) Eur. J. Biochem. 189,199–204.

9. Guan, K. L., and Dixon, J. E. (1991) Anal. Biochem. 192, 262–267.

0. Reiner, J. M. (1959) Behavior of Enzyme Systems, Burgess,Minneapolis.

1. Yang, J. T., Wu, C. S., and Martinez, H. M. (1986) MethodsEnzymol. 130, 208–269.

2. Bradford, M. M. (1976) Anal. Biochem. 72, 248–254.3. Laemmli, U. K. (1970) Nature 227, 680–685.4. Tan, A. W. (1979) Biochim. Biophys. Acta 582, 543–547.5. Busch, B., Hofmann, F., Selzer, J., Munro, S., Jeckel, D., and

Aktoties, K. (1998) J. Biol. Chem. 273, 19566–19572.6. Wiggins, C. A. R., and Munro, S. (1998) Proc. Natl. Acad. Sci.

USA 95, 7945–7950.7. Jansson, A. M., Jensen, K. J., Meldal, M., Lomako, J., Lomako,

W. M., Olsen, C. E., and Bock, K. (1996) J. Chem. Soc. PerkinTrans. 1, 1001–1006.

8. Lim, K., Ho, J. X., Keeling, K., Gilliland, G. L., Ji, X., Ruker, F.,and Carter, D. C. (1994) Protein Sci. 3, 2233–2244.