291Biochem. J. (1975) 152, 291-302Printed in Great Britain

Studies on the Glycosylation of Hydroxylysine Residues during Collagen Biosynthesisand the Subcellular Localization of Collagen Galactosyltransferase and Collagen

Glucosyltransferase in Tendon and Cartilage Cells

By RICHARD HARWOOD, MICHAEL E. GRANT and DAVID S. JACKSONDepartment ofMedical Biochemistry, Medical School, University ofManchester,

Oxford Road, Manchester M13 9PT, U.K.

(Received 4 June 1975)

1. The glycosylation ofhydroxylysine during the biosynthesis ofprocollagen by embryonicchick tendon and cartilage cells was examined. When free andmembrane-boundribosomesisolated from cells labelled for 4min with ['4C]lysine were assayed for hydroxy['4C]lysineand hydroxy['4C]lysine glycosides, it was found that hydroxylation took place only onmembrane-bound ribosomes and that some synthesis of galactosylhydroxy['4C]lysineand glucosylgalactosylhydroxy[14C]lysine had occurred on the nascent peptides. 2. Assaysof subcellular fractions isolated from tendon and cartilage cells labelled for 2h with[I4C]lysine demonstrated that the glycosylation of procollagen polypeptides beganin the rough endoplasmic reticulum. 14C-labelled polypeptides present in the smoothendoplasmic reticulum and Golgi fractions were glycosylated to extents almost identicalwith the respective secreted procollagens. 3. Assays specific for collagenl galactosyltrans-ferase and collagen glucosyltransferase are described, using as substrate chemicallytreated bovine anterior-lens-capsule collagen. 4. When homogenates were assayed for thecollagen glycosyltransferase activities, addition of Triton X-100 (0.01 %, w/v) was foundto stimulate enzyme activities by up to 45 %, suggesting that the enzymes were probablymembrane-bound. 5. Assays ofsubcellular fractions obtained by differential centrifugationfor collagen galactosyltransferase activity indicated the specific activity to be highest inthe microsomal fractions. Similar results were obtained for collagen glucosyltransferaseactivity. 6. When submicrosomal fractions obtained by discontinuous-sucrose-density-gradient-centrifugation procedures were assayed for these enzymic activities, the collagengalactosyltransferase was found to be distributed in the approximate ratio 7:3 betweenrough and smooth endoplasmic reticulum of both cell types. Similar determinations ofcollagen glucosyltransferase indicated a distribution in the approximate ratio 3:2 betweenrough and smooth microsomal fractions. 7. Assays ofsubcellular fractions for the plasma-membrane marker 5'-nucleotidase revealed a distribution markedly different from thedistributions obtained for the collagen glycosyltransferase. 8. The studies described heredemonstrate that glycosylation occurs early in the intracellular processing of procollagenpolypeptides rather than at the plasma membrane, as was previously suggested.

Analyses of collagens from various tissues demon-strated the existence of four genetically distinctcollagen molecules, which have been designatedTypes I-IV (for a review, see Miller, 1973). Chemicalanalyses of these collagenis demonstrated the presenceof galactose and glucose in covalent linkage withhydroxylysine, either as galactosyl-O-/i-hydroxyly-sine or 2-O-a-D-glucosylgalactosylhydroxylysine (fora review, see Spiro, 1972). The amount ofglycosylatedhydroxylysine found in different tissue collagensvaries from approx. one glycosylated hydroxylysineper 1000 amino acid residues in tendon and scleralcollagen to over 30 glycosylated hydroxylysineresidues per 1000 arnino acid residues in basement-membrane collagens (Spiro, 1969; Kefalides, 1973;Vol. 152

Grant, 1975). Theroles ofthese carbohydrate moietieshavenot been specifically defined, but the glycosylatedhydroxylysines may be involved in fibrillogenesis(Morgan et al., 1970), and they have also been impli-cated in the processes of collagen-mediated plateletaggregation (Barber & Jamieson, 1971a,b; Bosmann,1971; Chesney et al., 1972; Katzman et al., 1973;Kang et al., 1974).The synthesis of these hydroxylysine glycosides

has been shown to involve two enzymes, collagenUDP-galactosyltransferase and collagen UDP-gluco-syltransferase, both of which require Mn2+ (Bosmann& Eylar, 1968a,b; M. J. Spiro & R. G. Spiro, 1971;R. G. Spiro & M. J. Spiro, 1971a). Evidence waspresented indicating the localization of these trans-

R. HARWOOD, M. E. GRANT AND D. S. JACKSON

ferases in the plasma membranes of various cell lines(Hagopian et a!., 1968; Bosmann, 1969), but theseanalyses were carried out with a simple assayprocedure in which the UDP-[14C]glycoside wasincorporated into phosphotungstic acid-precipitablematerial with sugar-depleted collagen as substrate.This methodology has been criticized on severalgrounds (M. J. Spiro & R. G. Spiro, 1971; R. G.Spiro & M. J. Spiro, 1971a), but when the subcellulardistribution of these enzymes was investigated byusing the specific assay procedures of R. G. Spiro &M. J. Spiro (1971b), the data confirmed that thesetransferases were membrane-bound but their preciselocation could not be assessed.We have demonstrated that in freshly isolated

tendon and cartilage cells the first post-translationalevents in collagen biosynthesis, i.e. hydroxylation ofappropriate proline and lysine residues, commencewhile the growing procollagen polypeptides are stillattached to ribosomes (Harwood etal., 1973, 1974a,c,1975a), and the two enzymes protocollagen prolinehydroxylase and protocollagen lysine hydroxylaseare located predominantly in the rough endoplasmicreticulum of these cells (Harwood et al., 1974c). Todetermine the subcellular site of glycosylation ofhydroxylysine residues during the biosynthesis ofType I and Type II procollagens, we have examinedthe extent of glycosylation of procollagen poly-peptides isolated from subcellular fractions of thesecell types labelled under various conditions. Inaddition, the specific assays for collagen galactosyl-transferase (M. J. Spiro & R. G. Spiro, 1971) andcollagen glucosyltransferase (R. G. Spiro & M. J.Spiro, 1971a) activities have been modified to decreasethe time involved in assaying the labelled glycosidicproducts, and these procedures have been used in thesubcellular localization of the two transferases.Preliminary reports of parts of this work have alreadybeen presented elsewhere (Harwood etal., 1975b,c).

Experimental

MaterialsSubstrates for the assays of collagen glycosyl-

transferases were derived from lens capsules dissectedfrom bovine eyes obtained from a local abattoirwithin 1-2h of slaughter. Pepsin (3x crystallized) and5'-ATP were purchased from Signa (London)Chemical Co., London S.W.6., U.K. Ion-exchangeresin AG 50W-X8 (200-400 mesh) obtained fromBio-Rad Laboratories, St. Albans, Herts., U.K., wascleaned and prepared in the H+ form by the method ofMoore & Stein (1951) just before use. UDP-(U-14C]-galactose (300mCi/mmol) and UDP-[U-14C]glucose(260mCi/mmol) were obtained from the Radio-chemical Centre, Amersham, Bucks., U.K. Standardsof glucosylgalactosylhydroxy['4C]lysine and galacto-sylhydroxy['4C]lysine were prepared by ion-exchange

chromatography (Askenasi & Kefalides, 1972) of analkaline hydrolysate of procollagen secreted byembryonic chick tendon cells labelled for 2h with['4C]lysine as described below. All othermaterials andreagents were from sources described previously(Harwood et al., 1974b,c, 1975a).

Isolation and incubation ofmatrix-free cellsCells were isolated from the leg tendons and sterna

of 17-day chick embryos by digestion with trypsinand partially purified collagenase under conditionssimilar to those described by Dehm & Prockop (1971,1972, 1973). In the short-term pulse-labelling experi-ments cells were incubated at 37°C at a concentrationof 5 x 107 cells/ml of modified Krebs medium (Dehm& Prockop, 1971), but when cells were incubated for2h the concentration was 107 cells/ml ofmedium. Theincubations were conducted in silicone-treated glassErlenmeyer flasks, and cells were preincubated at37°C for 15min before addition of the appropriate'4C-labelled amino acid. Incubations were terminatedby rapid cooling to 4°C and the addition of cyclo-heximide to a final concentration of 100,ug/ml.Subcellular fractionation of cells

Preparation offreeandmembrane-boundcytoplasmicribosomes. Cells labelled for 4min with ['4C]lysinewere collected by centrifugation at 1200g for 5min.The cells were resuspended in 0.25 M-Sucrose in 0.05M-Tris-HCI buffer, pH7.5, containing 25mM-KCland 0.5mM-MgCI2, and homogenized under condi-tions described previously'(Harwood et al., 1974c).Nuclei and mitochondria were sedimented at 4°C bycentrifugation at 100OOg (ray. 6.5cm) for 10min inthe 10x lOml titanium angle rotor of an MSESuperspeed 65 ultracentrifuge. The pellet wasre-homogenized, then centrifuged at 100OOg for10min, and the two supernatants were combined.Free and membrane-bound ribosomes were obtainedby centrifugation on a discontinuous sucrose gradientas in the method of Blobel & Potter (1967).

Preparation of subcellular fractions by differenztialcentrifugation. Subcellular fractions were obtainedby procedures based on the method of Schachteret a!. (1970). Cells were homogenized in the sucrose-Tris-KCl-MgCI2 buffer as described above and thehomogenate was centrifuged at 600g for 5min in anMSE Mistral 6L centrifuge to sediment nuclei andcell debris. The nuclear pellet was resuspended in the0.25M-sucrose buffered solution and re-homogenized.This homogenate was resedimented at 600g for 5minto yield the nuclear fraction (N) and a supernatantwhich was pooled with the first postnuclear super-natant. The pooled supernatants were centrifugedat 10000g (ra,. 6.5cm) for 10min in the lOxlOmltitanium angle rotor of an MSE Superspeed 65centrifuge to yield a pellet and a supernatant (S3). Thepellet was resuspended in 0.34M-sucrose in 0.05M-

1975

292

GLYCOSYLATION OF COLLAGEN

Tris-HCI buffer, pH7.5, containing 25mM-KCI and0.5mM-MgCl2 and recentrifuged at 100OOg for10min to yield a mitochondrial fraction (ML) anda Golgi-enriched supernatant. This supernatant wasadjusted to 0.25M-sucrose and centrifuged at 10000gfor 10min to pellet the Golgi-enriched fraction (G).Supernatant (S3) was centrifuged at 1050OOg (ray.6.5 cm) for 1 h to obtain a microsomal pellet (M) andthe cytosol fraction (C).

Preparation ofsubmicrosomal fractions by discon-tinuous gradient centrifugation. Rough endoplasmicreticulum was separated from smooth endoplasmicreticulum by using a discontinuous-sucrose-density-gradient technique based on the method of Dallner(1963). Details of these centrifugation procedures andthe characterization of the fractions obtained havebeen described previously (Harwood et al., 1974c).

Analysis of subcellular fractions. Fractions wereassayed for the plasma-membrane marker, 5'-nucleotidase, by measuring the release of Pi from5'-AMP (Heppel & Hilmoe, 1951). This assay wasconducted under optimum conditions, and conditionsunder which product formation was proportional totime of incubation and to the quantity of enzymeprotein placed in the incubation mixture.

Protein was determined by the method of Lowryet a. (1951), with bovine serum albumin as standard.

Assay for hydroxy['4C]lysine glycosides by ion-exchange chromatography

Cells were incubated with ['4C]lysine for either4min or 2h and the extents of glycosylation ofhydroxy['4C]lysine in secreted procollagen, intra-cellular procollagen and procollagen polypeptidespresent in various subcellular fractions were deter-mined. Samples were dialysed against running tapwater for 18h and alkaline hydrolysis and subsequentdesalting were carried out as described by Grantet al. (1972). ['4C]Iysine, hydroxy('4C]lysine andhydroxy[l4C]lysine glycosides were separated by amodification of the method of Askenasi & Kefalides(1972). The hydrolysates were dissolved in 0.1 M-HCIand chromatographed at pH 5.3 on the long columnof a JEOL JLC 6AH amino acid analyser. Fractions(2.5 ml) were collected and samples (1 ml) werecounted for radioactivity in 10ml of scintillant asdescribed previously (Harwood et al., 1974b).

Determination ofcollagen glycosyltransferasesPreparation of substrates. Anterior-lens-capsule

collagen was prepared by a procedure based on themethod of Kefalides & Denduchis (1969). Anteriorlens capsules dissected from adult bovine eyes werefreed of cellular material by sonication in water fora total of ten 30s periods at 4°C at auto setting 6 onan MSE Sonicator. Capsules were freeze-dried andsubsequently digestedwith crystalline pepsin (enzyme/substrate ratio of 1:10) in 250 volumes of 0.1 M-acetic

Vol. 152

acid at 20°C for 18 h. The enzymic digest was centri-fuged at 40000g for 1 h and the supernatant wasdialysed for 18h against 0.01 M-acetic acid (8 litres) at4°C. To the sample was added solid KCI to a finalconcentration of 15% (w/v) and solid Na2HPO4 to aconcentration of 0.02M. The precipitated anterior-lens-capsule collagen was collected by centrifugationat 40000g for 1 h, redissolved in 0.01 M-acetic acidand dialysed against 0.01 M-acetic acid as above. Thecollagen preparation was then freeze-dried.The substrate for collagen galactosyltransferase

was prepared by removing the glucosylgalactosedisaccharide units on the anterior-lens-capsulecollagen by a single Smith degradation involvingperiodate oxidation, NaBH4 reduction and mild acidhydrolysis (Spiro, 1967). Approx. 100mg of freeze-dried collagen was incubated in 20ml of 0.02M-sodium metaperiodate in 0.05M-sodium acetatebuffer, pH4.5, at 4°C in the dark for 27h. Then 0.1 M-ethylene glycol (4ml) was added and the mixture leftfor 30min at 4°C before exhaustive dialysis againstwater. The non-diffusible material was suspendedin 15ml of 0.1 M-sodium borate buffer, pH 8.0, towhich was added 2.5ml of 1.0M-NaBH4 and thereduction was allowed to proceed for 24h at 4°C.The reaction mixture was adjusted top H5.0 withacetic acid, exhaustively dialysed against water,freeze-dried and then subjected to mild acid hydrolysisin 0.1 M-HCI at 80'C for 2h. The sample was neutral-ized with 2M-NaOH and used directly in the galacto-syltransferase assay.The substrate for collagen glucosyltransferase was

prepared by the selective removal of glucose fromthe basement-membrane collagen by hydrolysis in0.05M-H2SO4 for 20h at 100°C (R. G. Spiro & M. J.Spiro, 1971a). After hydrolysis the sample wasneutralized with 2M-NaOH and used directly in theassay.

Collagen galactosyltransferase assay. The incuba-tion was carried out in polypropylene tubes (5ml)for 2h at 37°C in a total volume of 200,ul, whichcontained l5umol of Tris-HCl buffer, pH 6.8,2,umol of MnCl2, 0.4p,imol of 2-mercaptoethanol,10ul of 0.1 % (v/v) Triton X-100, 0.63 pCi of UDP-['4C]galactose and 3mg of disaccharide-free lens-capsule collagen. The reaction was stopped by rapidlycooling to 4°C followed by the addition of 15 vol. ofacetone. The protein precipitate was collected bycentrifugation and the pellet washed a further threetimes each with 3 ml ofcold acetone. Alkaline hydroly-sis was carried out by adding 2ml of 2M-NaOH to thepellet in the polypropylene tubes, which were thensealed within large glass tubes and heated at 105°Cfor 18h.

After hydrolysis the samples were diluted withwater (lOml) and titrated at 4°C to pH 3.5-4.0 with1 M-HCI. The samples were applied to small columnscontaining 6g of Bio-Rad AG 50W-X8 ion-exchange

293

R. HARWOOD, M. E. GRANT AND D. S. JACKSON

resin (H+ form). The columns were washed withwater (50ml) and the hydroxylysine['4C]glycosideswere eluted with 1.5M-NH3 (25ml). The sampleswere dried by rotary evaporation of the NH3 at 40°Cand final traces were removed by freeze-drying afterredissolving the sample in water (4.0ml). The sampleswere dissolved in 1 ml of0.1 M-pyridine-acetate buffer,pH5.0, and 100,l portions were applied to 4cmx25cm strips of Whatman no. 1 paper and electro-phoresis was carried out in 0.1M-pyridine-acetatebuffer, pH5.0, at 250V for 4.5h. Standards ofgalactose, UDP-[(4C]galactose, galactosylhydroxy-I'4C]lysine, hydroxylysine and lysine were electro-phoresed on an adjacent strip. After drying at 100°C,the strips were cut into 1 cm sections and their radio-activity was determined in 10ml of toluene scintilla-tion fluid [5g of 2,5-diphenyloxazole and 0.3g of1,4-bis-(4-methyl-5-phenyloxazol-2-yl)benzene in 1litre of toluene].

Collagen glucosyltransferase assay. The incubationwas carried out for 2h at 37°C in a total volume of200u1 which contained 15,umol of Tris-HCI buffer,pH6.8, 5,umol of MnCl2, 0.4pumol of 2-mercapto-ethanol, 10#1 of 0.1 % Triton X-100, 0.67pCi ofUDP-(14C]glucose and 3mg of glucose-free lens-capsule collagen. Precipitation of the protein withacetone and subsequent alkaline hydrolysis, desalting,and electrophoresis of the samples were carried out asdescribed for the collagen galactosyltransferaseassay.

Characterization of the products formed by theglycosyltransferases. Previous assays for these en-zymes (Bosmann & Eylar, 1968a,b) are subject to thecriticism that the products of the enzymic reactionswere not fully characterized. The need to measurespecifically the hydroxylysine[14C]glycosides wasrecognized by R. G. Spiro & M. J. Spiro (1971a), whointroduced a paper-chromatographic separationwhich required 5-6 days to run. Such a time-consuming assay presents difficulties if it is to be usedin a multi-step purification procedure, and with thislong-term aim in view it was considered desirableto decrease the assay time without loss of specificity.By using defined substrates of high molecular

weight it was possible to precipitate the 14C-labelledproducts of the enzymic reactions with acetone andremove the exogenous UDP-[14C]glycoside. Of theradioactivity incorporated, not all was present ashydroxylysine[(4C]glycoside, since a proportion(30 %) ofthe 14Cwas not retained by the cationic resin.The major portion of the radioactivity eluted byNH3electrophoresed in a position corresponding tostandards of galactosylhydroxylysine and glucosyl-galactosylhydroxylysine, although a small peak ofradioactivity having a mobility intermediate betweenUDP-['4C]galactose and galactosylhydroxylysinewas usually observed (Figs. la and lb).

It should be noted that the electrophoretic proce-

dure does not separate the [(4C]glucosylgalactosylhydroxylysine from the ['4C]galactosylhydroxylysine,but the identities of these labelled species synthesizedin the respective enzyme assays with UDP-['4C]-glucose or UDP-[(4C]galactose were confirmed byion-exchange chromatography (Fig. 2).

Results

Analysis of procollagen polypeptides in subcellularfractions

In an attempt to establish the site of glycosylationof the hydroxylysine residues of procollagen, sub-cellular fractions were isolated from cells labelled

2.8

2.4

1.6

0.

* 2.

2.

o I.-d

x

0

8 k (b)

4-

6 -

8-

6 3 5 0 115t Distance (cm)

Origin

I

UDP-[14CIGal Gal |14CJHyl Hyl

Fig. 1. Paper electrophoresis ofthe products ofthe enzymicrelations

After alkaline hydrolysis, neutralization and desaltingof the acetoneprecipitable products of the enzymereactions, a portion was applied to Whatman no. 1

chromatography paper (3cm x 25cm) and electrophoresiswas carried out in 0.1 M-pyridine-acetate buffer, pH5.0, at250V for 4.5h. After drying at 100°C the strips were cutinto cm sections and the radioactivity was determinedas described in the text. (a) Products of galactosyltrans-ferase; (b) products of glucosyltransferase. The positionsof UDP-['4C]galactose,galactosylhydroxy[14C]lysine andhydroxylysine are shown on guide strips.

1975

(a)

294

--t

8-

nL

GLYCOSYLATION OF COLLAGEN

Gal-Hyl

Fcon20 30

Fraction no.

Fig. 2. Analysis by ion-exchange chromatography of theproducts ofthe enzyme reactions

Alkaline hydrolysates of the products of the enzymeincubations were neutralized and desalted then appliedto the long column of a JEOL JLC 6AH amino acidanalyser. Fractions (2.5ml) were collected and samples(1 ml) were counted for radioactivity in 10ml of scintillantas described previously (Harwood et al., 1974b). (a)Products of galactosyltransferase assay; (b) products ofglucosyltransferase assay; (c) Standards of glucosyl-galactosylhydroxy[14C]lysine, galactosylhydroxy[14C]-lysine, hydroxy[14C]lysine and ['4C]lysine prepared byalkaline hydrolysis of the procollagen present in themedium of embryonic-chick tendon cells incubated with[14C]Iysine for 2h at 37°C under conditions described in thetext.

with ['4C]lysine under steady-state conditions andanalysed for the presence of hydroxy['4C]lysine andhydroxy["4C]lysine glycosides. Initial analyses of theproteins secreted into the medium by tendon andcartilage cells incubated for 2h in modified Krebsmedium indicated that approx. 20 and 70% of thehydroxy["4C]lysine residues were substituted in therespective procollagens (Tables 1 and 2). Concomitantanalyses of the total cellular proteins revealed thatapprox. 15 and 60% of the hydroxy[14C]lysineresidues in intracellular tendon and cartilage pro-collagen polypeptides were glycosylated. These valuesindicate that the substitution of hydroxylysine hasreached over 70% of the final values observed for thesecreted molecules and suggest that in both cell types

Vol. 152

the intracellular pool of unglycosylated polypeptidesmust be relatively small and that the glycosylationprocess probably commences soon after hydroxyla-tion of appropriate lysine residues. Such a conclusionwould argue for a microsomal rather than a plasma-membrane location (Hagopian etal., 1968; Bosmann,1969) for the collagen glycosyltransferases, and thissuggestion is supported by analyses of subcellularfractions (Tables 1 and 2).When Golgi-enriched and microsomal fractions

prepared by the method of Schachter et al. (1970)were assayed for hydroxy[14C]lysine and hydroxy-[14C]lysine glycosides, the results indicated that thesynthesis of galactosylhydroxy['4C]lysine and gluco-sylgalactosylhydroxy[14C]lysine had commenced inthe microsomal fractions of both tendon andcartilage cells. The proportions of hydroxylysineresidues substituted with hexose in these microsomalfractions were approx. 50 and 80% of the valuesfound in the procollagens secreted by tendon andcartilage cells respectively. In contrast, the pro-collagen polypeptides present in the Golgi-enrichedfractions isolated from these cell types were glycosy-lated to an extent almost identical with the secretedmolecules.

Submicrosomal fractions of tendon and cartilagecells were isolated by the discontinuous-sucrose-density-gradient procedures of Dallner (1963).Analysis of these fractions revealed that procollagenpolypeptides present in the rough-endoplasmic-reticulum fractions contained galactosylhydroxy-[14C]lysine and glucosylgalactosylhydroxy[(4C]lysine(Tables 1 and 2). However, the ratios of disaccharideto monosaccharide were significantly lower than theratios determined for the extracellular procollagens,suggesting that the galactose and glucose residuesare not added simultaneously but in a stepwisemanner, which may be dictated by the spatial separa-tion of the glycosyltransferases. The procollagenpolypeptides in the smooth-endoplasmic-reticulumfractions were more extensively glycosylated, theproportions of free and substituted hydroxylysineand also the distribution between mono- and di-saccharide units being comparable with valuesobtained for the Golgi-enriched fractions isolated bydifferential centrifugation. This latter observationprobably reflects the fact that both cell types arelikely to have an extensive rough endoplasmic reti-culum and well developed Golgi apparatus, andtherefore separation of microsomes into rough andsmooth membranes might be expected to yieldsmooth-endoplasmic-reticulum fractions containinga large proportion of Golgi-derived vesicles.

Demonstration of the glycosylation of nascent pro-collagenpolypeptides on membrane-bound ribosomesThe demonstration that the synthesis of hydroxyly-

sine in cartilage cells occurs on nascent procollagen

295

R. HARWOOD, M. E. GRANT AND D. S. JACKSON

~4,C0

"C.)

'04)

CdON

U-.'

u0

CU

-'4

_.4

CUE

00.N%(~

e, _d ei .4 4

e0 teii

,

r, oce ct'.

00In to

env bt,

N000000

000 N

00 oo0 een 00 r.a e4

00

in@o88 8

oo 2

CU

A CU

.0

C.U00

00C.)

to

CU

a'.CU

.0

CU 4)U C

a'. Y-.C 4)

au C

S 0

CU

o.-CU

Ct 4-)

U)CU'0

X4) 0

0 0

c)

CU, D

Cr

'.'..0>CUC.)

CU,C>*-4 00C

.xo-

4)

0 r.

0 0-4)OCd

.0>4-4- o

Ca

4.4

C.)0

C>0

'02

4-

0

> p4

a',

1-NCU,

C.)

0

CU

0

(ONo -%No c0'r;oo en, e OeN

00.- -rt o

t-o t- NO

C4 en C4 en + N

0i 000%a Nt e

'd cmr'.6 e

ay 'fi 00m tn

-li C" 4 e-) Nene en 'R

.en

00 00 00

Ir) 000% 0No

Nen oo otn 00 '0 ON

Ira(NNt -'"oo en en (:O_C( t0 0

'04)

* o a,i en aso

- E°o °E) r ' - ct i a

_d

CU~~~~~^C

>4~~~~~C. Q

CU -

296

cU0

4)3

C) .

cUo

CU

Cd

CU.4.4.-

ON.

a"7

00 CU CUs

1975

GLYCOSYLATION OF COLLAGEN

Table 3. Analysis ofhydroxylation of lysine and glycosylation ofhydroxylysine in nascent procollagen polypeptides

Tendon cells (12.3 x 108) and cartilage cells (6.1 x 108) were incubated with 50pCi of [14C]lysine for 4min at 37°C. The incu-bation was stopped by cooling to 4°C and the addition ofcycloheximide. The cells were collected by centrifugation, washedand homogenized. Free and membrane-bound ribosomes were isolated and assayed for hydroxy[14C]lysine and hydroxy-[14C]lysine glycosides.

Free ribosomes

Total [14C]Hyl Gal-[14C]Hyl Glc-Ga(d.p.m.) (d.p.m.) (d.p.m.) (d;

Tendon 78900 194 31Cartilage 61000 183 36

Bound ribosomes

d1-[14CJHyl Total [14CJHyl Gal-[14C]Hyl Glc-Gal-[14C]HylL.p.m.) (d.p.m.) (d.p.m.) (d.p.m.) (d.p.m.)30 168700 14393 986 10334 106700 10713 1305 462

polypeptides attached to membrane-bound ribo-somes (Harwood et al., 1975a) and the above find-ings that the synthesis of hydroxylysine glycosidescommences in the rough endoplasmic reticulum ofboth tendon and cartilage cells prompted an investi-gation of the extent of glycosylation, if any, of thehydroxylysine residues of nascent polypeptides.Tendon and cartilage cells were labelled for 4minwith [14Cllysine, and free and membrane-boundribosomes were isolated by the method of Blobel &Potter (1967). That the 14C-labelled procollagenpeptides isolated after this labelling period representnascent chains was demonstrated in previous studies(Harwood et al., 1974a, 1975a) in which the pro-collagen peptides were found to sediment with largepolyribosomes on sucrose-density-gradient centri-fugation. Analyses of free and membrane-boundribosomes (Table 3) confirmed that the synthesisof tendon and cartilage procollagen occurs only onmembrane-bound ribosomes, and that the synthesisof hydroxylysine occurs on nascent tendon procolla-gen polypeptides also.The detection of galactosylhydroxy[14C]lysine

and glucosylgalactosylhydroxy[14C]lysine in themembrane-bound ribosomal fractions provides thefirst demonstration that the glycosylation of pro-collagen commences while the growing chains arestill attached to the ribosomes (Table 3). In thefraction from the tendon cells approx. 7% of thehydroxy[14C]lysine is substituted, of which less than10% is present as the disaccharide unit, whereas inthe cartilage sample approx. 14% ofthehydroxy[14C]-lysine is substituted, of which 25% is present asglucosylgalactosylhydroxy["4C]lysine.

Effect ofTriton X-100 andprotein concentration on theactivities ofthe collagen glycosyltransferasesTo investigate further the glycosylation ofhydroxy-

lysine residues, studies on the collagen galactosyl-transferase and collagen glucosyltransferase wereundertaken. Initial experiments using homogenatesof the matrix-free cells indicated that the presence of0.01 % Triton X-100 in the incubation mixture

Vol. 152

<. 12-(bS

48 _0 0.4 0.8

Cd~ ~ Prti (mg3

1'4 12 (bux

4

0' 0.4 0.8

Protein (mg)Fig. 3. Effect ofenzyme concentration onproductformation

in the enzymic reactions

Tendon (0) and cartilage (o) cells were homogenized in0.25M-sucrose in Tris-KCl-MgCl2 buffer, thehomogenatewas diluted with buffer to the equivalent of approx. 2.5 x106 cells/ml and samples of0.02,0.04, 0.06, 0.08 and 0.lmlwere assayed for collagen galactosyltransferase andcollagen glucosyltransferase as described in the text. (a)Collagen galactosyltransferase; (b), collagen glucosyl-transferase.

stimulated the activity of both glycosyltransferases byapprox. 45 %. The detergent was therefore added as aroutine to all subsequent assays, and (as shown inFig. 3) the activities of these enzymes derived fromboth tendon and cartilage cells were proportionalto the enzyme concentrations.

297

R. HARWOOD, M. E. GRANT AND D. S. JACKSON

~oen i tooC; C5

6,

14

_

00o^ m O 00 C

-oeq 0 f4

......4

C-4

_C oo _

C-q8o2kn0 ,a

0 .

CU X '_eWC

CU C UFO 44) we_£J114110

Distribution of collagen galactosyltransferase andcollagen glucosyltransferase in subcellularfractions oftendon and cartilage cells

Subcellular fractions obtained by the centrifugationschemes described above were assayed for proteincontent and both collagen galactosyltransferase andcollagen glucosyltransferase activities. The resultsshown in Table4 demonstrate that in both tendon andcartilage cells the distribution patterns of thegalactosyltransferase are almost identical, the majorportion of this enzyme activity being detected in themicrosomal fractions. In further experiments thegalactosyltransferase was shown to be distributedthroughout the endoplasmic reticulum, with approx.70% of the total activity being present in the roughmicrosomal fractions. The specific activities of theendoplasmic-reticulum fractions are lower than thoseof the microsomal fractions obtained by differentialcentrifugation. Similar observations were reportedin studies of the subcellular localization of proto-collagen proline hydroxylase and lysine hydroxylasein these cell types (Harwood et al., 1974c), and alsoby Schachter et al. (1970), who studied the subcellulardistribution of a variety of sugar transferases in ratliver. These losses of enzymic activity may beexplained by gradual denaturation during thediscontinuous - sucrose - density - gradient - centri-fugation procedure.

Table 5 reports the results obtained for the dis-tribution ofcollagen glucosyltransferase in subcellularfractions from both tendon and cartilage cells. Theresults show that the two transferases are similar intheir distribution patterns, but in this case slightlymore of the glucosyltransferase activity (approx.40t%) is associated with the smooth-endoplasmic-reticulum fractions.

Comparison ofthe distributions ofthe collagenglycosyl-transferases and the plasma-membrane marker,5'-nucleotidase

Since previous studies (Hagopian et al., 1968;Bosmann, 1969) have raised the possibility thatcollagen glycosyltransferases may be bound to theplasma membrane, it was considered important toestablish the fate of the plasma-membrane fragmentsin the centrifugation procedures used in the presentstudy. Subcellular fractions obtained by differentialcentrifugation and by sucrose-density-gradient centri-fugation were assayed for the enzyme 5'-nucleotidasewhich is generally considered to be a plasma-membrane marker and which has been shown to bepresent in plasma membranes isolated fromembryonic chick tendon cells (Lehtinen et al., 1975).In both centrifugation procedures the majority ofthe 5'-nucleotidase activity was found in the nuclearand mitochondrial fractions (Fig. 4 and Table 6).Less than 15% of the total activity is present in themicrosomal fraction, whereas this latter fraction

1975

*t

-4

A..

I-

,0CU

(.1

CU

a ;t4...1

cd

u)CU

cc

CUD0C

I.

14).cd00

0'0CU~

298

11

GLYCOSYLATION OF COLLAGEN

3.0o-000 o0 s N

0 o Wt o w0 _-

2.0

0o o N0 0 Noo o6o C6V

1.0.

.D 0 Nb 0 0 eICm tn "

el;. 0. *

N - 't 0 N O0r~ -NN-

6666666

N 00oo 0 000n In 000O.0-,

-o6ri6N

tnr-o oom C-4 C-4 C,

-;N6 (6 6000 1

- o 'N -

0%

-N N Noo N

e0N -41 N 00

te) N N N -

0-

- 0.C)

Ce

0o 0 EcC)Fugl

C.

._

C.

8

0.C)Ce)

0

3.0O

0 50 100Protein (Y.)

Fig. 4. Distribution of collagen glucosyltransferase and5'-nucleotidase in subcellularfractions obtained by differen-tial centrifugation oftendon and cartilage cell homogenates

The subcellular fractions are: N, nuclei; ML, mitochondriaand lysosomes; M, microsomal fraction; C, cytosol. Eachof the fractions was assayed for protein content, andthe collagen glucosyltransferase activity associated withthe fractions is presented as a hatched histogram super-imposed on the distribution of the 5'-nucleotidase activitypresent in these fractions. (a) Tendon cell fractions; (b)cartilage cell fractions.

contains over 55 % of the collagen glucosyltransferaseactivity in both tendon and cartilage cells (Figs. 4aand 4b). The small amounts of 5'-nucleotidaseactivity present in the microsomnal fractions are foundto be associated predominantly with the smooth-endoplasmic-reticulum fractions (Table 6), whereasthe major portion of the glycosyltransferase activitiesis located in the rough-endoplasmic-reticulumfractions (Tables 4 and 5).

299

C1)

Ce3

00

0

C-)

r.00 .- ,

.- !t (AE

r

CeC)

C)

CeZ

0

C).0

0.C

0 0

C)

2~Ce00

3.x4

C)

L.

C:

0

I.

3..

*0

C)-0

a

C)

0.

0

la0C)

Vol. 152

R. HARWOOD, M. E. GRANT AND D. S. JACKSON

Table 6. Distribution of5'-nucleotidase activity in submicrosomalfractions oftendon and cartilage cells

Tendon and cartilage cells were homogenized in 0.25M-sucrose and fractionated by the method of Dallner (1963). Fractionswere assayed for protein and 5'-nucleotidase activity and values are given relative to those obtained with the crude homo-genate.

Tendon

Protein 5'-Nucleotidase(%/) (%~)

Cartilage

Protein 5'-Nucleotidase(%/) (%~)

HomogenateNucleiMitochondriaRough endoplasmic reticulumSmooth endoplasmic reticulumCytosol

Discussion

During the last decade the processes involved inthe synthesis and secretion of glycoproteins werestudied in considerable detail and it is now generallyaccepted that secretory proteins are synthesized on

membrane-bound ribosomes and that carbohydrateis added as they are transported from the roughendoplasmic reticulum through the smooth endo-plasmic reticulum to the Golgi apparatus and out ofthe cell. Recent studies with matrix-free tendon andcartilage cells (Harwood et a!., 1973, 1974a,d, 1975a;Olsen et al., 1973; Olsen & Prockop, 1974) havedemonstrated that the synthesis and secretion ofprocollagen follows the general pathway of secretedglycoproteins, but the site of glycosylation of thehydroxylysine residues has not been satisfactorilydefined.

In glycoprotein biosynthesis the glycosylationprocess frequentlycommences either with the additionof xylose to peptidylserine or with the addition ofN-acetylglucosamine to peptidyl-asparagine, andfurther glycosylation occurs by the sequentialaddition of appropriate sugar residues catalysed bymembrane-bound glycosyltransferases located alongthe secretory pathway. The glycosylation of collagen,however, requires that the product of translation ofprocollagenmRNA undergoes prior hydroxylation ofpeptidyl-lysine residues by protocollagen lysinehydroxylase. This latter process is found to com-mence while nascent procollagen polypeptides arestill attached to the ribosomes (Table 3), and themolecule thus becomes a substrate for the collagengalactosyltransferase at the ribosomal level. It was

originally proposed that the subsequent addition ofgalactose and then glucose under the direction of thespecific collagen glycosyltransferases occurred justbefore secretion, since these enzymes were initiallyreported to be associated with the plasma membrane(Hagopian et al., 1968; Bosmann, 1969). However,the demonstration that glycosylation of hydroxyly-sine during the biosynthesis of basement-membrane

collagen occurred at an early stage in the intracellularprocessing of this procollagen (Grant etal., 1972) didnot appear to be consistent with this hypothesis.The ability to obtain well-characterized subcellular

fractions of tendon and cartilage cells (Harwoodetal., 1973, 1974c, 1975a) has enabled us tore-examinethe intracellular site of collagen glycosylation. Initialanalyses of subcellular fractions obtained by differen-tial centrifugation (Tables 1 and 2) indicated thatprocollagen polypeptides in the microsomal fractionwere substantially glycosylated, suggesting that theaddition of galactose and glucose commenced earlyin the intracellular transport of both tendon andcartilage procollagens. When submicrosomal frac-tions isolated by discontinuous-sucrose-density-gradient centrifugation were analysed, it was evidentthat polypeptides present in the rough-endoplasmic-reticulum fractions were partially glycosylated,whereas the extents of glycosylation of the poly-peptides in the smooth-endoplasmic-reticulum frac-tions were almost the same as those of the secretedprocollagens (Tables 1 and 2). These results were

obtained with cells labelled under steady-stateconditions and in further studies using short-termpulse-labelling conditions it was demonstrated thatthe addition of both galactose and glucose wasinitiated on growing peptide chains attached tomembrane-bound ribosomes (Table 3). Thus theglycosylation of procollagen appears to occur soon

after the synthesis of hydroxylysine, and it will benoted that the hydroxylysine of nascent cartilageprocollagen polypeptides is substituted to a greaterextent than in the tendon polypeptides, an observa-tion which may reflect the higher activities of theglycosyltransferases in the cartilage cells (Tables 4and 5).The above results are at variance with the sugges-

tion that the collagen glycosyltransferases are locatedon the plasma membrane (Hagopian et al., 1968;Bosmann, 1969). However, this latter suggestion wasbased on studies using assay systems which havesince been criticized for their lack of specificity

1975

10031.323.65.15.6

33.2

10052.730.62.19.34.2

10033.825.15.24.9

30.7

10056.128.33.08.72.9

300

GLYCOSYLATION OF COLLAGEN 301

(R. G. Spiro & M. J. Spiro, 1971a; M. J. Spiro &R. G. Spiro, 1971) and, in the case of the galactosyl-transferase, it has been pointed out (M. J. Spiro &R. G. Spiro, 1971) that the preparation of the sub-strate by using two successive periodate oxidationswould destroy any hydroxylysine-acceptor sites inthe protein. When the subcellular distribution oftheseenzymes was examined in rat renal-cortical tissue byusing assays developed to measure specifically thesynthesis of the hydroxylysine glycosides (R. G. Spiro& M. J. Spiro, 1971b), the enzymes were shown tobe present primarily in a rapidly sedimenting parti-culate fraction. This fraction was not characterizedfurther, although the observation that the enzymeactivities were increased by the presence of TritonX-100was in contrast with previous reports (Hagopianet al., 1968; Bosmann, 1969) and would be consistentwith a vesicular location for these enzymes.To determine the subcellular location of the colla-

gen glycosyltransferases in the tendon and cartilagecells, the specific assay procedures (R. G. Spiro &M. J. Spiro, 1971a; M. J. Spiro & R. G. Spiro, 1971)were modified to overcome the step involving thetime-consuming separation of the hydroxylysineglycosides by paper chromatography. By introducingthe paper-electrophoretic procedure described above,it has been possible to decrease the time involved bya factor of 3 without any loss of specificity. It shouldbe noted that during the preparation of this manu-scipt a similar modification was reported by Risteli &Kivirikko (1974).

Preliminary evidence that both the transferases aremembrane-bound in tendon and cartilage cells wasobtained by studies of cell homogenate assayed forthese enzymes in the presence and absence of 0.01 %Triton X-100. Both enzyme activites were found tobe stimulated (approx. 45 %) in the presence of thedetergent, confirming the previous observations ofR. G. Spiro & M. J. Spiro (1971b) and Risteli &Kivirikko (1974). Assays of the subcellular fractionsfor the two transferases indicated that less than 15%of the total activities were present in the cytosolfractions and the relative specific activities of bothgalactosyltransferase and glycosyltransferase werefound to be highest in the microsomal fractionsisolated from tendon and cartilage cells (Tables 4and 5). When submicrosomal fractions obtainedby discontinuous-sucrose-density-gradient-centri-fugation procedures were assayed for these enzymicactivities, the collagen galactosyltransferase wasfound to be distributed in the approx. ratio 7:3between rough and smooth endoplasmic reticulum ofboth cell types. Similar determinations of collagenglucosyltransferase indicated a distribution in theapproximate ratio 3:2 between rough and smoothmicrosomal fractions.

These studies appeared to eliminate the plasmamembrane as the site of the transferases. but the

possibility of plasma-membrane fractions con-taminating the microsomal fraction could not beeliminated entirely. However, preliminary investiga-tions of the influence of Triton-X100 on the glyco-syltransferases in cell homogenates had indicatedthat these enzymes exhibited structure-linkedlatency in that they were further activated by thedetergent (see above). The plasma-membrane marker5'-nucleotidase, has been shown not to exhibit suchlatency in homogenates of rat liver (De Duve, 1971)and rat embryo fibroblasts (Tulkens et al., 1974), andin studies of the activity of this enzyme in tendon- andcartilage-cell homogenates we were unable to detectany stimulation of 5'-nucleotidase by the addition ofTriton X-100. These results provided the first indica-tion that the glycosyltransferases and the 5'-nucleotid-ase had different locations and this observation wasconfirmed by the demonstration of a lack of coincid-ence between the distributions of 5'-nucleotidase andthe collagen glycosyltransferases in the subcellularfractions of both tendon and cartilage cells (Fig. 4and Table 6).The discovery that the collagen glycosyltransferases

are associated predominantly with the rough-micro-somal fraction is in contrast with the more usualobservations of glycosyltransferases being presentin smooth endoplasmic reticulum and Golgi fractions(Fleischer et al., 1969; Schachter et a!., 1970).However, several demonstrations that the initiationof the assembly of the carbohydrate side chains ofglycoproteins may be a ribosomal event havepreviously been reported (Molnar & Sy, 1967; Sherr& Uhr, 1970; Schenkein & Uhr, 1970; Uhr &Schenkein, 1970). In our earlier studies of thesynthesis of hydroxyproline and hydroxylysine intendon and cartilage cells (Harwood et al., 1973,1974a,c, 1975a) it was observed that these post-translational events commenced on nascent poly-peptides and it now appears that the subsequentaddition ofgalactose and glucose to the hydroxylysineresidues can also occur at the ribosomal level. Theseobservations suggest that the four enzymes respon-sible for the secondary modifications after thetranslation of procollagen mRNA may reside in acontiguous relationship as a multi-enzyme systembound to the internal face of the cisternae of the roughendoplasmic reticulum.

We thank Mr. A. Colwell and Mrs. Joan Ward forskilled technical assistance. The support for this study bygrants from the Science Research Council and MedicalResearch Council is gratefully acknowledged.

References

Askenasi, R. S. & Kefalides, N. A. (1972) Anal. Biochem.47, 67-72

Barber, A. J. & Jamieson, G. A. (1971a) Biochim. Biophys.Acta 252, 533-545

Vol. 152

302 R. HARWOOD, M. E. GRANT AND D. S. JACKSON

Barber, A. J. & Jamieson, G. A. (1971b) Biochim. Biophys.Acta 252, 546-552

Blobel, G, & Potter, V. R. (1967) J. Mol. Bio. 26, 279-292

Bosmann, H. B. (1969) Life Sci. 8, 737-746Bosmann, H. B. (1971) Biochem. Biophys. Res. Commun.

43, 1118-1126Bosmann, H. B. & Eylar, E. H. (1968a) Biochem. Biophys.

Res. Commun. 30,89-94Bosmann, H. B. & Eylar, E. H. (1968b) Biochem. Biophys.

Res. Conmwn. 33, 340-346Chesney, C. M., Harper, E. & Colman, R. W. (1972)

J. Clin. Invest. 51, 2693-2701Dallner, G. (1963) Acta Pathol. Microbiol. Scand. SuppL.

166, 1-94De Duve, C. (1971) J. Cell Biol. 50, 20D-55DDehm, P. & Prockop, D. J. (1971) Biochim. Biophys.Acda 240, 358-369

Dehrm, P. & Prockop, D. J. (1972) Biochim. Biophys. Acta264,375-382

Dehm, P. & Prockop, D. J. (1973) Eur. J. Biochem. 35,159-166

Fleischer, B., Fleischer, S. & Ozawa, H. (1969)J. CellBiol.43,59-68

Grant, M. E. (1975) Symp. Soc. Study of Inborn ErrorsofMetabolism 11th in the press

Grant, M. E., Kefalides, N. A. & Prockop, D. J. (1972)J. Bio. Chem. 247, 3539-3544

Hagopian, A., Bosmann, H. B. & Eylar, E. H. (1968)Arch. Biochem. Biophys. 18, 387-396

Harwood, R., Grant, M. E. & Jackson, D. S. (1973)Biochem. Biophys. Res. Commun. 55, 1188-1196

Harwood, R., Connolly, A. D., Grant, M. E. & Jackson,D. S. (1974a) FEBS Lett. 41, 85-88

Harwood, R., Grant, M. E. & Jackson, D. S. (1974b)Biochem. J. 142, 641-651

Harwood, R., Grant, M. E. & Jackson, D. S. (1974c)Biochem. J. 144, 123-130

Harwood, R., Grant, M. E. & Jackson, D. S. (1974d)Biochem. Biophys. Res. Commun. 59, 947-954

Harwood, R., Bhalla, A. K., Grant, M. E. & Jackson,D. S. (1975a) Biochem. J. 148, 129-138

Harwood, R., Grant, M. E. & Jackson, D. S. (1975b)Protides Biol. Fluids. Proc. Colloq. 22, 83-85

Harwood, R., Grant, M. E. & Jackson, D. S. (1975c)Biochem. Soc. Trans. 3, 136-137

Heppel, L. A. & Hilmoe, R. J. (1951) J. Biol. Chem. 188,665-676

Kang, A. H., Beachey, E. H. & Katzman, R. L. (1974)J. Biol. Chem. 249, 1054-1059

Katzman, R. L., Kang, A. H. & Beachey, E. H. (1973)Science 181, 670-672

Kefalides, N. A. (1973) Int. Rev. Connect. Tissue Res. 6,63-104

Kefalides, N. A. & Denduchis, B. (1969) Biochemistry 8,4613-4621

Lehtinen, P., Vuorio, E. & Kulonen, E. (1975) Biochem. J..146, 565-573

Lowry, 0. H., Rosebrough, N. J., Farr, A. L. & Randall,R. J. (1951) J. Biol. Chem. 193, 265-275

Miller, E. J. (1973) Clin. Orthop. Rel. Res. 92, 260-281Molnar, J. & Sy, D. (1967) Biochemistry 6, 1941-1947Moore, S. & Stein, W. H. (1951) J. Biol. Chem. 192,663-670

Morgan, P. H., Jacobs, H. G., Segrest, J. P. & Cunning-ham, L. W. (1970) J. Biol. Chem. 245, 5042-5048

Olsen, B. R. & Prockop, D. J. (1974) Proc. Natl. Acad.Sci. U.S.A. 71, 2033-2037

Olsen, B. R., Berg, R. A., Kishida, Y. & Prockop, D. J.(1973) Science 182, 825-827

Risteli, J. & Kivirikko, K. 1. (1974) Biochem. J. 144,115-122

Schachter, H., Jabbal, I., Hudgin, R. L., Pinteric, L.,McGuire, E. J. & Roseman, S. (1970) J. Biol. Chem.245,1090-1100

Schenkein, I. & Uhr, J. W. (1970) J. Cell Biol. 46,42-53Sherr, C. J. & Uhr, J. W. (1970) Proc. Natl. Acad. Sci.

U.S.A. 66, 1183-1188Spiro, M. J. & Spiro, R. G. (1971) J. Biol. Chem. 246,49104918

Spiro, R. G. (1967) J. Biol. Chem. 242, 4813-4823Spiro, R. G. (1969) J. Biol. Chem. 244, 602-612Spiro, R. G. (1972) in Glycoproteins (Gottschalk, A., ed.),2nd edn., pp. 964-989, Elsevier, Amsterdam

Spiro, R. G. & Spiro, M. J. (1971a) J. Biol. Chem. 246,48994909

Spiro, R. G. & Spiro, M. J. (1971b) J. Biol. Chem. 246,49194925

Tulkens, P., Beaufay, H. & Trouet, A. (1974) J. Cell Biol.63, 383 401

Uhr, J. W. & Schenkein, I. (1970) Proc. Natl. Acad. Sci.U.S.A. 66,952-958

1975