interaction of a brain extracellular matrix protein with hyaluronic acid

248 Biochimiea et Biophpsica Acid. 11)75 (1991) 248-258 .~) 1991 Elsevier Science Publishers B.V. All rights resetwed 1131}4-4165/91/$03.511

ADONIS" 0304416591tR12523

BBAGEN 23597

Interaction of a brain extracellular matrix protein with hyaluronic acid

G e o r g e P e r i d e s , F i l i p p o B i v i a n o a n d A m i c o B i g n a m i

Department of Pathohrgy, Harl'ard Medical School attd Spinal Cord Injury Research Laboratory. Department of Veterans Affairs Medical Center West Roxbury, MA (U.S.A.)

(Received 2 April 1991)

Key words: Extracellular matrix; Glia; Hyaluronate binding protein

A glial hyaluronate.binding protein (GHAP) was isolated from bovine spinal cord and partially characterized. Bovine GHAP consisted of three immunologically related polypeptides with mogecular masses of 76, 64, and 54 kDa and isoelectric points of 4.1, 4.2, and 4.4, respectively. Peptide mapping and partial amino acid sequencing showed that all three polypeptides derive from the same protein. The protein was localized immunohistochemically with rabbit antisera in the white matter surrounding the myelinated axons. Sugar analyses indicated that the three polypeptides are glycosylated and the sugar residues account for at least 30% of their weight. After enMmatie deglycosylation, the apparent molecular mass of the bovine (]HAP was reduced to 43 kDa. The biochemical properties of bovine GHAP were compared to those of human GHAP. Initial peptide mapping indicated similarities between bovine and human GHAP. Partial amino acid sequencing of .bovine GHAP showed a striking identity (up to 90%) with human GHAP and with the hyaluronate binding domain of the large human flbroblast proteoglycan, versican. Bovine and human GHAP were demonstrated to bind specifically to hyaluronic acid (HA) with one protein molecule binding to an average 17 disaccharide repeating units. The binding of bovine and human GHAP was inhibited by oligosaccharides of HA and specifically by the octamer. Salt concentrations of up to 1 M NaCI had very little effect on the binding of the GItAP to HA. The GHAP-HA interaction was pH dependent. Dissociation only took place at low pH ( < 3.5). Analysis of several polypeptides derived from GHAP by limited proteolysis allowed us to conclude that one of the tandem repeated sequences is sufficient for HA binding and that the aminotermina! domain (which contains an immunoglobulin-like fold) is not involved in the GHAP-HA-binding event.

Introduction

Electron microscopy studies in the late 1950s indicated that the extracellular space in the central nervous system (CNS) is extremely small, accounting for less than 5% of the brain volume [1]. It is now gener- ally accepted, however, that the apparent size of brain extracellular space in these electron micrographic studies of conventionally processed tissue is an artefact due to cellular swelling resulting from asphyxia. More re- cent data suggest that the real extracellular space in

Abbreviations: HA, hyaluronic acid; GHAP, glial hyaluronate-binding protein.

Correspondence: G. Perides. Department of Pathology, Harvard Medical School and Spinal Cord Injury Research Laboratory, De- partment of Veterans Afhdrs Medical Center, West Roxbu~, MA 02132, U.S.A.

mature brain occupies 17-20% in volume [2,3]. Little is known with regard to the content of this space although brain extracellular matrix may be important of studies of CNS regeneration.

Recently we reported the isolation of a glial hyaluronate-binding protein (GHAP) from human brain white matter obtained at autopsy [4]. Human GHAP is a 60 kDa glycoprotein that appears to share a high degree of similarity with other extracellular proteins and specifically with those of cartilage; proteoglycan and link protein [4]. The identity demonstrated between the partial amino acid sequences of human GHAP and the primary sequence deduced from the eDNA sequence of the hyaluronate-binding region of human versican, a large fibroblast proteoglycan, could suggest that GHAP originates from the versican by limited proteolysis [5]. The precise relations between GHAP and hyaluronectin, a hyaluronate binding protein also isolated from brain tissue [6] still remain to be determined, but it should be noted that the localization

of the two proteins is completely different [7]. In the present study, we investigate the specificity of the binding of GHAP to hyaluronie acid (HA) and the inhibition of the binding by short oligomers of HA and other glycosaminoglycans. It will be shown that the HA-binding properties of GHAP are different from those of cartilage proteoglycans and link protein ai~d of the hyaluronate receptor. Furthermore, we identify the region of the protein that is important for the binding to HA. lr. order to avoid post-mortem autolysis, this study was conducted on bovine GHAP.

Materia ls and Methods

Endoproteinase Lys-C, endoproteinase Arg-C. Irypsin and neuraminidase were purchased from Boehringer Mannheim. Staphylococcal VS proteinase and bovine testes hyaluronidase were from Miles. N- glycanase and O-glycanase were obtained from Gen- zyme Corporation. Concanavalin A, bovine serum albumin aad Y,3'-diaminobenzidine were from Sigma. Goat anti-mouse and goat anti-rabbit rbodamine and peroxidase conjugated antibodies were from Cooper Biomedical. The 9 \ 3 0 \ 8 - A - 4 monoclonal antibody to link protein [8] and the 1 2 \ 2 l \ 1-C-6 monoclonal antibody to rat chondro~rcoma proteoglycan monomer [9] were obtained as ascites fluid from the Developmental Studies Hybridoma Bank. Ampholytes, molecular weight and pl standards were from Pharmacia-LKB Biotcchnolo~. Polyvinylidene difluoride (lmmobilon) membranes were purchased from Millipore. [~H]HA, 0.22 m g / m l at a 2.46. l0 s cpm//~g specific activity was kindly provided by Dr. C.B. Underhill and prepared as described previously [10]. All other chemicals were of analytical grade and obtained either from Sigma or from J.T. Baker Chemicals.

Analytical methods Sugar quantitative analyses were performed accord-

ing to the method described by Spiro [11] and David- son [12]. Fucose was determined as described by Dis- che and Sbettles [13], galactose and mannose after hydrolysis in 0.5 M H2SO 4 according to Park and Johnson [14], sialie acids according to Warren [15] and glucosamine and galactosamine through differential colorimetry after hydrolysis as described by Levvy and McAllan [16] and Elson and Morgan [171. Hyaluronic acid determination was made after treatment with hyaluronidas¢ according to Elson and Morgan [17]. Protein determination was performed according to Lowry et al. [18].

Preparation of antibodies Polyclonal antibodies to bovine GHAP were pre-

pared in one female New Zealand white rabbit, 3.5 kg

,% 4d,~

in weigh,. The rabbit was injected subcutaneously witi-, 400 /.tg of electrophoretically purified GHAP emulsi- fied in complete Freund's adjuvant followed 3t) days later by am,ther subcutaneous injection of 200 /xg of antigen in incomplete Freund's adjuvant. The rabbit was bled frora the car I0 days later, and the serum was tested for immunoblotling and immunofluorcscenee against GHAP. Absorption experiments of the monoclonal and polyclonal antibodies with bovine GHAP were performed as described previously [4].

Preparatiotts of oligosaccharides from hyahtronic acid Oligosaccharides from HA were prepared based on

the method described by Weissmann et al. [19]. Briefly. 20 m g / m l of HA was digested in 0.2 M NaCH3COO (pH 5.0) and 150 mM NaCI at 3 7 ° C with 5 m g / m l of bovine testes hyaluronidase. The digestion products were concentrated through an Amicon filter with a 10000 kDa cut-off. The oligomers were applied to a DEAE-Sepharose CL6B in 20 mM phosphate buffer (pH 7.0) and eluted with a 0 to 0.3 M NaCI gradient. The fractions were analysed by thin-layer chromatography on silica gel using isopropanol/H20 (66/34) containing 50 mM NaCI as a solvent system and the oli,.omers were visualized using 10% CuSO~ in 8% H~POa and incubating at 120°C for 30 min.

Electrophoretic procedures Sodium dodecyl-sulfate polyacrylamide gel elec-

trophoresis (SDS-PAGE) was performed according to Laemmli [20], and two-dimensional electrophorcsis was canicd out in the manner described by Cells and Bravo [21], For p l determination on two-dimensional electrophoresis, the p l calibration kit of Pharmacia (3.5- 9.7) was used. Staining of the proteins in SDS gels was performed with 1% Coommassie brilliant blue in 25% methanol, 10% acetic acid and destaining was with 10% acetic acid. Silver staining was performed according to Morissey [22]. Electrophoretically purified protein was obtained by elution from destained SDS- PAGE gels using a Bio-Rad 422 electrceluter. lmmunoblotting was performed after elcctrophoretic transfer of proteins to nitrocellulose as reported previously [23]. Primary antibodies for immunoblotting were used at a 1 : 500 dilution in Tris-buffered saline (TBS) (pH 7.5) containing 0.05% Tween 20, and secondary antibodies at a 1:2000 dilution. Visualization was achievcd with Y,3'-diaminobenzidine.

lmmunofluorescence Bovine spinal cord, cerebellum, and cerebral cortex

were dissected within 15 min after death and cryostat sections were fixed in acetone at - 2 0 ° C prior to staining with the antibodies a,* a ! : 20 dilu,'~on in TBS, (pH 7.5) containing 0.1% bovine serum albumin [4].

250

[:nz.vmatic clearage Cleavage of bovine and human GHAP with the

Lys-C and Arg-C specific cndoproteinases was performed in TBS (pH 7.5) at 37°C. Proteolytic degradation with trypsin was carried out in TBS (pH 7.5) with 0.1 mM Ca z÷ at 1 : 100 enzyme to substrate ratio at 37 ° C. Cleavage with V8 proteinase was performed in 75 mM NH4CI-13COO (pH 5.0), 4 mM EDTA at 3 7 ° C and at a 1 : 10 enzyme to substrate ratio.

Binding of GtlAP to hyahtronic acid Binding of GHAP tn l-!.~ was perform..e,J .9¢¢or,_t.ir!g

to Underhill et al. [24]. Briefly, 20 p.I of 0.5 m g / m l bovine GHAP was incubated with increasing amounts of [~H]HA in TBS (pH 7.5) at room temperature for 15 min in a final volume of 2130 ~1. Protein was precipi- tated by the addiiion of equal volumes of 100% saturated (N H4)zSO ~, centrifuged 8 0 0 0 x g for 5 min, washed with 400 p.l of 50% saturated (NH4)2SO 4 and then centrifuged once more. Combined supernatant and pellet were counted in an ICN Taurus automated liquid scintillation cocktail counter.

Amino acid analysis and sequencblg Amino acid analysis was performed on elec-

tropboretically purified protein transferred to lmmobilon [25] in an Appl ied Biosystems 420 A derivatizer/130 A analyzer for amino acid analysis. For amino acid sequencing proteolytic products of bovine GHAP were transferred to Immobilon stained and destained as described previously [25]. iodividual bands were cut and washed extensively with H , O and subjected to analysis by automated Edman degradation on an Applied Biosystems, Inc. 470A gas-phase sequenator equipped with an on-line 120A phenylthiohydan- ruin analyzer. All sequenator reagents and solvents were from Applied Biosystems.

Results

Isolation of but'me GHAP :solation of bovine GHAP was performed essentially

by the method we have reported for the human protein [4]. Bovine GHAP was extracted from bovine spinal cord by homogenization (pH 2.8) and affinity purified on an HA-Sepharose column [26]. The protein content at each stage was monitored by immunoblot with a polyclonal antibody raised ag=inst human GHAP [4]. "i he incubation at 0 °C for 30 rain at pH 5.6 appeared to be essential, since immediate centrifugation after adjusting the pH and increasing the salt concentration to 150 mM led ;o sedimentation of the protein and very low yields. In an average of nine different preparations, 5.2 mg of purified protein were isolated from 100 g of bovine spinal cord. Unlike the human GHAP, which migrated as a single polypeptide of 60 kDa in

SDS-PAGE, the bovine species apparently consisted of three polypeptides with molecular masses of 76,64, and 54 kDa. The 64 kDa polypeptide was consistently the main species present and accounted for at least 75% of the total amount of isolated protein. In different prt:parations the amount of the other two species appeared to vary.

Lower molecular weight contaminants when present could easily be removed by subjecting the protein extracts to CM-Sepharose fast-flow chromatography in 6 M urea and 10 mM NaCH3COO (pH 5.0), during which the three po!ypcptides o_!uted at 30 mM Nat l . Alternatively, the protein can be applied to a DEAE- Sepharose fast-flow (pH 7.6) and eluted at 240 mM NaCI. Care should be takeil in all steps to avoid any reducing agents, such as/3-mercaptoethanol or dithio- threitol since the ability of GHAP to bind onto HA decreases dramatically after t reatment with reducing agents in the presence of 6 M urea despite their apparent removal through extensive dialyses (d~:ta not shown), in some preparations a mixture of protease inhibitors (l mM phenylmethylsulfonic fluoride, 1 mM 1-tosylamido-2-phenylethylchloromethylketone i mM leupepfin and I mM EDTA) was present in the ho- mogenate, but no difference in the molecular mass of the isolated proteins was observed. In SDS extracts and in the presence of these proteinase inhibitors immunological identification of the 76 and the 64 kDa proteins was consistently possible. Identical results were obtained when GHAP was extracted by homogenization in the presence of 0.5 M guanidine-HCl and proteinase inhibitors [27]. The 54 kDa polypeptide however was observed only after the extraction at pH 2.8 (not shown). Although rabbit antibodies were raised against bovine GHAP using only electrophoretically purified 64 kDa protein, all three polypeptides were immunoreactive with this antiserum. We obtained the same results with the polyclonal antibodies raised against human GHAP, and the three monoclonal antibodies, 12D6, 12C5 and 6F7 [28], which we currently use in this laboratory. Absorption experiments were performed as described in the Materials Methods, and both human and bovine GHAP were equally able to prevent the staining of the GHAP antigen on immunoblots. Neither the 9 / 3 0 / 8 - A - 4 monoclonal antibody specific for the link protein and the hyaluronate binding region of cartilage proteoglycans [8] nor the 12 /21 /1 -C-6 specific for the core proteoglycan [9], reacted with the proteins on immunoblots (data no shown).

Localization To localize the protein in the ccntral nervous sys-

tem, cryostat sections from bovine spinal cord, cerebellum and cerebrum were stained by indirect immuno- fluorescence. Most GHAP immunoreactlvity was found

251

~ " TABLE I

3"ugar ~,mpo,itUm of hob'm,, and human (;/IAP

, , • Numhcr~ ~how I:ercentages (w /w) and are the average tff flint • diffclcnt dctcrminatinns from two different hydrnlyscs

Suga~ Bovine GHAP Iluman GI IAP

~ ~ - ~3h _~,~ (;alacto~,c 6.45 7.50 Manno:,c 5.80 5.70 N-e\celylgl aco~lm ne 2.13 1.20 V-Ace lylga,aclo.~lm nc 4.17 218l) Sialic acids 7.99 6.1gl

Fig. I. Localization of bovine GHAF in bovine spinal cord. Trans- verse sections of bovine spinal cord were stained with the rabbit antiserum raised against bovine GtlAP and visualized with a goat anti-rabbit rhodamine-conjugated antibody. The irnmunoreactive material forms a mesh surrounding individual myelinated axoas. Glial s~pta (arrows in a) and glial limitans on Ihe surface of the

anterior fissure (arrows in b) are also stained. Bar, 12.5.

in the white matter. With the exception of the granular layer in the cerebellum, gray matter stained only faintly. On transverse sections of spinal cord the immuno-fluo- rescent staining appeared as a mesh surrounding the myelinated axons (Fig. 1). Glial septa (arrows, Fig. la) were stained, as well as the glial limitans on the surface of the spinal cord (apparent at higher magnification in Fig. Ib). The same immunofluore~ent s:aining pattern for bovine tissue was obtained using each of the monoclonal and polyclonal antibodies we previously raised against human GHAP (data not shown).

Characterization The p l of thc thrcc bovine GHAP polypeptides was

4.1, 4.2 and 4.4 for the 76, the 64 and 54 kDa pol)peptide, respectively (Fig. 2). We reported a similar p l (4.3) for human GHAP. Sugar analyses of the bovine and human GHAP were next performed as described in the Materials and Methods (Table It. Bovine GHAP contains more sugar residues than the human protein

and this proved consistent with the larger molecular weight reduction obtained through sequential deglycosylation with N-glycanase, ncuraminidase and O- glycanase [4]. After incubation with N-glycanase, the apparent molecular mass of the major species de- creased from 64 to 54 kDa; after incubation with neuraminidase, to 52 kDa and after incubation with O-glycanase to approx, 43 kDa (Fig, 3). After incubation with O-glycanase the bands appeared very diffuse on SDS-PAGE suggesting that complete removal of the sugars could not be achieved (Fig. 3, lane 4). Binding of concanavalin A to the protein was completely abolished aftcr incubation with N-glycanasc indicating that all mannose residues were removed (not shown). The incubation with N-glycanasc and O -

g l y c a n a s e did not have any effect on the isoelectric points of the proteins. After treatment with neuraminidase, however, all three polypeptides appear to shift to a more basic form (4.7) (not shown). Deglycosy- lation did not have any apparent effect on the immuno- reactivity of the bovine GHAP polypeptides since our anti-GHAP polyclonal antibodies bound equally well to the deglycosylated protein (data not shown).

Enzymatic digestion Like its human counterpart, the bovine protein ap-

peared to be resistant to proteolysis so that relatively large amounts of enzymes were used for proteolytie

- - . . _ - . . . . . . . ÷

.d , I

Fig. 2. Two-dimensional electrophoresis of GHAP. (a) Silver slaincd gel of 5 pg of bovine GHAP subjected to isoelectrofocusing in a 200-~.I micropipetle with ampholytes 3.5-10 [19] and 12.5c7c SDS-PAGE. (b) The same as ,, with the addition of 5 p.g of human GHAP (arrowhead).

252

digestion. SDS-PAGE analyses revealed a number of stable pept idcs (Fig. 4), which were used fur ther for amino acid scqucncing atudics (see below). Pept ide maps af~.er l imited proteolysls were i'irst compared to the dige';tion products we have repor ted for human GHAP. With the exception of the pept ides ob ta ined after cleavage with the Lys-C specific endoorote inase , the bovine pept ide digests were very s imilar to those of the human protein demons t ra t ing only a slight shift to a higher molecular weight (compare with Fig. 8 in Ref. 4). With highe, com.~ntrat ions of the Lys-C specific endopro te inase only the 42 and the 36 kDa polypept ides were present , a pa t te rn that was identical to the digest ion pa t te rn of human G H A P [4]. Trypsin digestion first gave rise to a 31 kDa polypept ide which was apparen t at 30 min, and then to a 29 kDa polypeptide which was the only form present af ter 4 h of incubation (Fig. 4, panel c).

Amino acM analysis and sequencing Amino acid analysis was pre formed on elec-

t rophoret ical ly purif ied protein (64 kDa) t ransferred to Immobi lon as descr ibed in the Mater ia l s and Methods.

- M_I. 2 3 .,1 M _

P,i . -

+ + ,+ +, +,+.. ....

Fig. 3. Sequential dcglycosylation of GHAP. 100/zg of bovine GHAP polypeplides were incubated in 0.2 M phosphate buffer (pH 8.6). 100 mM/3-mercaptoethanol at 37 o C fo= 16 h with 2 units of N-glyeanase (lane 2). The pH was adjusted to 6.1 with H~PO 4 and GHAP was incubated for 2 h at 370C with 0.2 units of ncuraminidase (lane 3). 0.02 unhs of O-glycanase were added and incubated for 6 h at 37°C (lane 4). After each step, 2 p.g of protein were subjected to SDS. PAGE and silver stained• Lane t, GHAP control. M, molecular weight standards. From top to bottom: phosphor'/lase b, 97000; bovine serum albumin. 6?000; ovalbumin. 43000; carbonic anhy- drase, 30000; trypsin inhibitor, 20100; a-lactalbumin, 14400.

M 1 2 3 4 5 6 7 8 M M 1 2 3 4 5 t / 7 8

Ill

i i ++ + ++~:? -

~ -!:+++:.,. ,

:~ % . I . 1 . g a 4 5 0 Z 8 M

• ~ . !

. . . . .: -atom= -

i m l l l l u ' .

, , mm~ , - mmmN--

" " ~ m _ D - (ram,

m m m q m ~ m ~ m

c d Fag. 4. Enzymatic digestion of OHAP. 100 p.g of bovine GHAP were run in a 10-15% SDS-PAGE and silver stained following enzymatic degradation at 37 o C with 3 units of endoproteinase Arg-C (panel a), 0.03 units of endopmteinase Lys-C (panel b), I #g of [wpsin (panel c), and 10 ,ag of V8 proteinase (panel d). Aliquots were taken after 0 rain (lane 2), after l0 rain, (lane 3), after 30 rain (lanc 4). after (10 rain (lane 5), after 2 h (lane 6), after 4 h (lane 7), after 16 h (lane 8). and after 44 h (lane 9). Lane I, GHAP control. M, molecular weight

standards (as in Fig. 3).

The amino acid composi t ion of the bovine G H A P was s imilar to human G H A P and the bovine nasal car t i lage l ink prote in (Table I1). In o rde r to more careful ly compare these two proteins , we sequenced a number of the proteolytic p roducts d isplayed in Fig, 4. The 29 kDa polypept ide ob ta ined with tryptic d iges t ion (T 29), the 21 kDa polypept ide produced with cleavage with V8 prote inase (S 21), and the 54,42,36,29,25, and 18 kDa polypept ides ob ta ined af ter incubat ion with the Lys-C specific endopro te inase were used (L 54, L 42, L 36, L 29, L 25, L 18). The following sequences were obta ined:

T 29: S P P V K G S L S G K V N L P C H F S T M P - L P S 21: L O K V * M E K S P P V K G S L S G K V N L P L 18: D L K E T I ' V L V A Q D G N 1 K I G D

TABLE It

Compari~u, o f tile amino u('id ~'O,ll;O.~ttion of hol the altd human (;HAP [4L boring" nasal ¢'artihtge Iml~ prott'i, (tlN('Lp) 120/. rut chotldrosclrcoma Iblk protcitl 2 (R('I.P2) /30]. and ('hWl~c, li,k tlrot,.t,i /('LP~ 13/I

Numbers show percentage, and bovine (;tlAP i,, lhe average of l~.~ur determinations from two different preparations. Ar, p is the ",urn of asparlic acid and asparagine: Glu is the ~tlrll of glulamic acid and glulamine. ('y~lcine and t~'ptophan v.ere no! determined

Amino acid h-(_it lAP h-(.;I lAP BN('LP R('I.P2 ('LP

A~,p 'L I I 1.0 13.2 12. I ~.¢~ Glu I11. l) 111.7 S.2 N~ IliA} Ser 5.6 7.2 6,.5 511 6.5 Gly 11.4 9.1 1115 II1.11 7.3 llis 1.5 2.') 2.7 2.7 31 Arg 5.6 6.4 6. I 7.2 ~5 T h r 5.9 6.2 5.5 4.4 3.7 Ala 8.1 ~.S 7.9 7.t~ 6.~ Pro 5.5 7111 5.2 5 .l) 4 5 T y r 3.6 _.'~ ¢, 6.0 6. I 5. I val s.I s.t h i ~ 7 c Met 1.0 0.h I).3 11.4 I .I) lie ~ 0 3.3 2.S 3.7 4.,": Lcu 7~ S.I ,'1.2 S5 S.2 Phc 3.1 3.5 5.11 5.8 4.2 l.ys h. i 4.8 5.8 5.4 5.6

L 25: S P P V K . . .

L 29: E G V R T Y G F R A P H E T Y D V Y C Y V D H L L 36: E G V R T Y G F R A P H E T Y D V Y C Y V D H L D

253

L 42: D L K E T T V L V A O L 54: D L K . . .

Asterisk indicates possible position for glycosylation. The S 21 polypeptide is close but not identical to the

N- terminus of versican repor ted previously [5] and overlaps with the T 29 and L 25 polypeptides. Polypep- tide L 29 appears to be der ived from the peptidc f ragment L 36. Polypeptidc L 18 appears to come from polypcptidc L 42 which first originates from L 54. Compar ison with the sequences we have already repor ted from the human G H A P showed that the two

proteins are highly similar suggesting that the few amino atcid differences are species related (Fig. 5). Compar ison with the hyaluronate binding domain of human versican [5] also indicates a very high degree of identity ( f rom 73 to 90%) (Fig. 5~. In an a t tempt to identify, the amino acid sequence of the C-terminus of the protein the L 36 polypeptide was digested in situ

wlil~ t,ypsin and the der ived polypeptides were sequenced [32]:

L 36-NT28: FEN L 36-NT67: A Q C G G G L L G V R L 36-NT84: C D Y G W L L D A S V R L 36-NT88: F D A Y R F K

L 36-NT92: F T F E E A G E E C K ' I Q D A R

All sequences share a high degree of identity with

a 5 2 1 T ~

' L Q X V . N [ K S P P V K G S L S G K V N L P C H r S T N P . L p i i i i ! /

V = I L H K V K V G K S P P V R G S L S G K V S L P C H F S T H P T L P s a

X y K V G K S P P V R G S L S G ~ S L P C H ; S T H P

S4 TZS

b L 18

T T D L K ( T T V L V A Q N N Z K Z G D Y K G R V S V P T H P E A V I I 0

E T T V L V A G nl N ZK T G Q O Y K G R V S V P T H P E A V

$ 9

¢ L ] l

II [ 6 V R T Y S F R A P H E T Y D V Y C Y V O H L D I

v ,,IIA G V It T Y G F R S Pl~ E T Y 0 V Y C y V O H L 0 GD V FH L T V p $ ig F,, ,

H T Y G F I S P O E T Y D V Y C Y V O H L n G O V F H L T V P $ I R

T 10 T 31

Fig. 5. Comparison of amino acid sequences of bovine GHAP with human GHAP and human versican. (a) Comparison of the combined sequences S 21 and T 29 with the combined sequences S 4 and T ~ of human GHAP (H) (Peridcs el al., Rcf. 4) (erratum corrige: sequence S 9 of human GI'IAP should read E'ITV instead of EETV as published) and vcrsican IV) (Zimmermann and Ruoslahti, E.. Ref. 5). (b) Comparison of the sequence L 18 with I-| and V. (c) Comparison of the sequence L 36 with El and V. Non-identical amino acids have been connected.

Numbers indicate the position of the amino acids in the pr in l a l~ ~|I'UCItI~C Of ;'Ci";iC~.

254

M 1 2 . . ~ , ~ _ 6

i'" !llI) .. g

= . , I I) 11 - I ~ - - r

Fig. 6. Limited trjptic digestion of the 76 and 54 kDa polypeptides. Silver stained gel of limited digestion of electrophoretieally purified 76 and 54 kDa polypeptidcs. Lane I. isolated 76 kDa polypeptide species: lane 2. isolated 76 kDa species and trypsin at time 0; lane 3. degradation products after 1 h at 37°C; lane 4, isolated 54 kDa species: lane 5. isolated 54 kDa species and trypsin at time 0; and lane 6. degradation products after incubation at 37 ° C. M. molecular

weight standards as in (Fig. 3).

versican and the L 36-NT88 appears to extend till the 347 amino acid position of the versican.

In order to determine if the two minor species of the bovine GHAP are separate proteins or derive from the same parent protein, the two polypeptides were electrophoretically purified and limited proteolytic digestion with trypsin wac performed on them as well (Fig. 6). The peptide maps were very similar to the one obtained from the main (64 kDa) species (compare with Fig. 4c). The 76 kDa species revealed two major polypeptides with apparent molecular masses of 40 and 31 kDa. The 54 kDa species also revealed two major polypeptides with apparent molecular masses of 34 and 31 kDa.

The 31 kDa polypeptides were transferred to lmmobilon and sequenced. Both were found to have the same sequence as the S 21 and the T 29 polypeptide from the main species of 64 kDa ( L Q K V * . . . ) . The limited proteolysis was performed under conditions unfavorable for tryptic digestion (50 mM Tris-HCl (pH 8.8), 400 mM glycine and 150 mM NaCI). It was necessary to avoid tryptic digestion in TBS with 0.l mM CaLl2, since electrophoretically purified protein proved to be denatured and more susceptible to proteolysis with trypsin than the proteins from the original preparation and v;'e could not obtain stable polypeptides. The zlectrophoretic mobility of )he 31 kDa polypeptide from tryptic digestion of 76 and 54 kDa polypeptides was not reduced to 29 kDa even after 16 h incubation

with trypsin. Indeed, the amino terminal sequence of the 31 kDa polypeptide is identical to the amino terminal sequence of the S 21 polypeptide which starts at the amino terminus of the versican and. eight amino acids earlier than the T 29 polypeptide.

Binding onto hyaluronic acid Binding of bovine GHAP to HA was performed as

described by Underhill et al. [24]. GHAP bound increasing quantities of HA in a saturable manner (Fig. 7a). Maximum binding was reached within I min. in Fig. 7b, the line starts above 0 indicating the back- ground binding which was consistently observed to be between 5.2-6.3% of the [3H]HA offered in the incubation mixture. Fig. 7b, which describes the amount of [~H]HA bound as a function of increasing amounts of GHAP demonstrates that a standard amount of about 103 ng HA could be absorbed per # g of protein (Fig. 7b) which is similar to the amount of [aHIHA bound human GHAP (113 ng//.tg of protein). The binding of the [-~H]HA was displaced by the addition of increasing amounts of non-radioactive HA (Fig. 7c). Despite the similarity of human and bovine GHAP with the link protein we could not inhibit the binding with the 9 \ 3 0 \ 8 - A - 4 monoclonal antibody raised against the HA binding region of rat link protein which cross-re- acts with the human link protein [8] and inhibits the binding of chicken link protein to HA [33]. In addition, none of our monoclonal antibodies (6F7, 12C5 and 12D6) had any effect on the ability of GHAP to bind to HA. The pH dependency was investigated by incubating GHAP-HA complexes at different pH values. GHAP was stably bound to HA across a very wide range of pH values. At pH 3.5 about 75% of the [3H]HA was bound compared to the control at pH 7.5. However, at pH 3.0 only 15% of the binding ability was retaine6 (Fig. 7d). This is consistent with the dissociation at low pH of cartilage proteoglycans bound to HA [34,35] and probably explains why GHAP is best extracted from the brain and the spinal cord at low pH. To investigate the dependency of binding on ionic interactiors, GHAP was incubated with HA at different NaCI concentrations, in the presence of 500 mM NaCI the GHAP binding of [3H]HA was about 95% of the control at NaCI concentration of 150 raM. At 2.5 M NaCI bovine GHAP demonstrated about 55% of its [3H]HA binding capacity (Fig. 7e). To determine further the stability of the binding interaction, we incubated [3H]HA-GHAP complexes in different concentrations of NaCI for 15 min. Concentrations of up to 1.0 M NaCI did not appear to have any effect on the stability, while higher concentrations lead to dissociation of the [3H]HA-GHAP complexes (Fig. 7e).

Specificity of the binding In order to investigate further the specificity of the

binding of GHAP to HA we used short oligomcrs o[

D-glucuronic acid and N-acetyl-o-glycosamine. Differ- ent rlttios of tetra-, hexa-, octa-, and decasacchar ide to H A were allowed to compete fi~r bin, ' ! g to G H A P . At higher ratios of ol igomers to H A (10t)t): 1 and IiH): 1)

each of the ol igomers appea red to compete at different s t rengths for binding to G H A P . At smaller ratios ( Ill : l and I : 1) only the oc t amer anti the decamer of H A appea red to compete for the binding (Fig. 8). In fact, at a I : 1 ratio the oc t amer inhibits the binding of [~H]HA to bovine and human G H A P by 51) and 53c; -. respectively. At a 1 : l ratio the d ecamer inhibits the binding less effectively than the detainer. On it weight basis, the same amount of decanter would result in ;i 2IV,: lower

255

concentrat ion than the oc tamer and consequently to 211c; less molecules able to compete for binding. Hep- arun sulfate, chondroit in sulfate A, chondroit in sulfate B (de rma tan sulfate) and chondroit in sulfate C at a

lllll: I ratio did not have any effect on the ability of the G I I A P to bind to H A (not shown).

Binding of GtlAP-derived polypeptide.s to HA In order to de te rmine the domains of the protein

part icipate in H A binding we per fo rmed limited proteolytic digestion of bovine and human G H A P s~.ith the

V8 proteinase, trypsin, the Lys-C and the Arg-C specific cndoprote inase . The V8 degradat ion products (S

a

0

S /

2 4

(3HI-HA offered (POt

O

10

0.8

06

i 04

0.2

0 0 - - 2 4 6 8 1o 12

GHAP effered tpg)

12 c

t o

08

06

04.

0.2

o.o t0 20 30 40 50

HA offilnl,d (pg)

d

100 O

!6o ~ 40

02 4 6 8 to

tool

8O

6O'

40"

20"

O.0 0.5 1 0 1.5 Z0 2.5

pet [NICJ] (M)

Fig. 7. Bindm8 of GHAP to hyaluronic acid (a) tO Pg of GHAP were incubated with increasing amounts of [3HIHA. Bound [:~FI]HA was determined as descril~d under Materials and ~lethocls. Bound radioactivity was correlated to the amount offered of ['~H]HA to the mixture. (b) 2 /zg of ['~HJHA were incubated with increasing amounts of (3HAP and the measured radioactivity was Correlated to the amount ol GHAP offered to the mixture. (c) 2 p.g of [ '~tIIHA were incubated with t0/.~g of GHAP in the presence of increasing amounts of non-radioactivc HA and the radioactivity measured was correlated [~, the i~on-radioactlve HA offered to the incubation mixture. (d) 2.5 pg of bovine GHAP were incubated for 15 rain with 0.285 v.g of [;H]HA and the pl-| was adjusted with the addition of "fris-NaCH.~COO buffer to 2.5 till 10.5 and the amount of radioactivity found in tile pellet was correlated to the amount bound at pH 7.5. (e) 2.5 g.g of bovine GHAP were incubated with 0.285 t.tg of ['~I-I]tIA for 15 min and then the NaCI concentration was increased up to 2.5 M and the bound radioactivity was corrclatcd to the amount found after incubation at 150 mM (O). 2.5 v.g of bovine GI-IAP were incubated with 0.285 of [~H]HA for 15 minutes in the presence of

increasing NaCI concentrations and the amount bound to GHAP was correlated to the amount bound at 150 mM NaCI (12).

256

a b

100 - t 00 ] I ~

o~ 80 80

"~c is1 o = 60 80 ~ i o , ,

. 40 40 [ ] I o c ~ , I

20 20

0 0 4 6 8 10 4 6 8 10

length of oligosaccharldes length of ollogoaacchartdes

Fig. 8. Inhibition of binding of GHAP to HA by oligosaccharides. 2.5 p.g of bovine (panel a) and human GHAP (panel b) were incubated with 0.285 gg of ['~HIHA in the prcscncc of tetra-, hcxa-, octa-, and decasaccharide at a I000/1. 100/I. 10/1 and t / I oligosaccharide/['~H]HA

(w/w) ratio and the bound radioactivity was compared to the amount bound in the absence of oligosaccharides (1(]0%).

21 polypept ide) , the A r g - C specif ic e n d o p r o t e i n a s e p r o d u c t A 31 a n d the t rypsin p r o d u c t T 29 (see above Fig. 4, a n d Ref. 4) d id not a p p e a r to have any b ind ing ability. Interes t ingly, e a c h o f these po lypep t ides beg ins at, o r close to the a m i n o t e rminus o f the p ro t e in a n d con ta ins the immunog lobu l in - l ike fold a c c o r d i n g to the s econda ry s t r uc tu r e p r e d i c t e d f rom the p r i m a r y se- q u e n c e o f vers ican [5] (Fig. 9). T h e L 42, L 36, L 29 a n d L 18 we re all f o u n d to b ind to [ ~ H ] H A a n d to H A immobi l ized to Sepharose . Desp i t e o u r effor ts , however, with reverse p h a s e h igh -p re s su re l iquid ch ro - m a t o g r a p h y a n d gel f i l t ra t ion in the p re sence o f 4 M guan id ine HCI, we were no t able to s e p a r a t e the vari-

NH2 T . . . . . COOH I

T29 Fig. 9. Positioning in the HA-binding region of the various poiy- pcptides derived from bovine and human GHAP. The various polypeptides tested for their ability to bind onto HA were positioned on the seconda~ structure of the hyaiuronatc binding region of rat chondrosarcoma link protein [33], rat cartilage proteoglycans core protein [36], versican [SJ as predicted by their complete eDNA sequence and the partial sequences of human, chicken, pig and bovine aggregating proteoglycans and link proteins I37]. The exact position of the polypeptides was based on the primary structure of the vcrsican. The dashed lines indicate disulfide bonds between cysteines. The V 21 and the T 31 start at the amino terminus of the versican. The L 42 contains part of the immunoglobulin-like fold and probably extends to both the tandem repeat sequences. The L 36 and L 29 derived from bovine GHAP and the L 28 derived from human GHAP contain only a small part of the first and the second tandem

repeat sequence.

pus polypeptides obtained after cleavage with the Lys-C specific endoproteinase. The L 28 fragment from human GHAP was also prepared using a 2 h incubation

Fig. 10. Isolation of L 28 polypcpBd¢ from human GHAP. 100/z 8 of human GHAP were incubated in the presence of 150 mM NaCI in 10 mM Tris acetate (pH 7.5) for 2 h at 37°C with 0.02 units ol Lys-C specific endoproteinase. The degradation products were through an HA-Scpharose column and bound polypcptides were elated with 4 M guanidine HCI in 50 mM NH4CH3COO (pH 5.8). The guanidine HCI was removed by dialysis against H:O and material correspond. ing to f pg of human GIIAP (lane 1) and the L 28 polypcptide (lane 2) were run on a 15% SDS-PAGE and silver stained. M, molecular

weight standards (as in Fig. 3).

with the Lys-C specific endoproteinase (Fig. I l l ) . Tilt,, polypeptidc proved to be identical to the L 2~.) from the bovine OHA P with an aminotcrminal sequence ( A G V R T . . . ) which corresponds to the portion of versican starting at amino acid 225 [5]. The binding was found to be inhibited at a 10: I (oligt;iner/FIA) ratio by the octamer and the detainer but not the hexamer. The L 28 polypeptide from human GHAP starts at the 225th amino acid of the versican amino-tcrminus sequence and contains the second of the two tandem repeat sequences (Fig. 9).

Discussion

We have now purified bovine GHAP using affinity chromatography of spinal cord extracts on an HA-Scp- harose column. Unlike its human counterpart which consists of only one protein band, bovine GHAP appeared as three bands on SDS-PAGE (76. 64 and 54 kDa). A yield of 5.2 mg of protein per 100 g of wet tissue could be obtained, which is significantly less than the yield from human cerebral white matter (,~.2 rag/100 g wet tissue), probably because the protein was isolated from spinal cord that contains a fair amount of gray matter. After enzymatic deglycosylation, the three polypeptides were shown to appear as one wide band and partial amino acid sequencing of tryptic digestion products showed that the original three polypeptides derive from the same protein. Whether the three polypeptides differ in the sugar content a n d / o r in the amino acid sequence, remains to be seen.

Comparison of bovine G H A P with human GHAP by tissue localization, amino acid composition, peptide mapping and partial amino acid sequencing shows that they are homologous proteins. Furthermore, both proteins appear similar to the HA-binding region of the cartilage proteins, proteoglycan and link protein. Re- cently, the complete eDNA derived amino acid sequence of the large human fibroblast proteoglycan, versican, was published [5]. The partial antino acid sequences from human GHAP, reported earlier [4] were virtually identical to the sequences within the HA-binding region of versican. The sequences reported in this publication allow us to conclude that the same applies to the bovine protein.

The similarity of GHAP with the HA-binding region of versican raises the possibility that G t tAP is derived from the large (265 kDa) proteoglyean versican by post-mortem proteolysis, since the human GHAP was first isolated from human autoptic material [4]. The finding that bovine G H A P consists of similar molecular weight polypeptides is not in favor of this possibility. In addition, it appears unlikely that proteolysis occurred during the isolation procedure since we isolated proteins which had identical molecular weight whether or

257

not protein:,se inhibitors were present during the isola- lion steps. It is also possible that GHAP originates lrom versican during the initial extraction procedure with H('I. Acid sensitive peptide bonds in vcrsican arc identifiable at positions 375-376 and 436-43"/ 151. However. the acid sensitive peptide bond,; between aspartic acid and proline, require 24-96 h incubation at 11) ° C in the presence of denaturants and pH below 2.5 to achieve breakage [38]. Given that our conditions included low temperature (2 ° C) and a short extraction duration (for a maximum time of 1 h) this possibility is unlikely. In addition, we identified proteins by immunoblotting with the same molecular weight in SDS and guanidine HCI-extracts of I~wine spinal cord. The band pattern had remained essentially unchanged even when the extraction was conducted on spinal cords kept for 4 h at 20 C, thus indicating that GHAP is resistant to post-mortcm autolysis.

B,winc GHAP binds specifically to HA with a binding capacity of 103 ng of HA per l /.tg of protein. This means that each molecule of protein occupies approxi- mately the length of 17 disaccharide repeating units. However. we also noted specific inhibition of binding to GHAP by the octasaccharide. The decamer oligosaccharide specifically inhibits the binding of hyaluronate to cartilage protcoglyeans and link protein [391 and the hexamer inhibits the binding to the hyaluronate receptor of the plasma membrane of fi- broblasts [24]. None of the other glyeosaminoglycans tested inhibits the binding. High salt concentrations also have little effect on the stability of the binding. The GHAP-HA complexes dissociate only at very low pH (under 3.5) which explains the recovery of GHAP from the affinity chromatography column following 10 mM elution with 10 mM HCI [41.

The tryptic and V8 proteinase degradation products of bovine and human GHAP did not bind to HA. Both the T 29 and S 21 polypeptides contain the inamuno- globulin-like fold (Fig. 9). The degradation products from the cleavage with the Lys-C specific endoproteinase, however, bound to HA. In addition, the L 36, L 29 polypeptides from bovine GHAP and the L 28 polypcptide from human GHAP, which do not contain the immunoglobulin-like fold but do contain the second of the tandem repeat sequences (Fig. 9) were found to bind to HA. This allows us to conclude that as reported by Goetinck et al. [33] and despite previous indications to the contrary [40] the immunoglobulin-like fold is not necessary for the binding to HA. tt is interesting to note, that only one of the tandem repeat sequences is contained in CD44, a lymphocyte homing receptor [41,42]. It was recently reported that CD44 is the principal cell surface receptor for hyaluronate [43].

The localization of GHAP in the extracellular space of dog spinal cord white matter has been recently demonstrated by immunoelectron m~croseopy and it

258

has also recently been shown that G H A P binds hyaluronate in vivo [28]. We conclude, based on these data, that for the first t ime an extracellular matrix protein has been identified in ma tu re CNS white mat-

tcr. It is interest ing to note that inhibition of neural crest cell migrat ion of fibronectin by a chondroit in sulfate proteoglycan is media ted by its hyaluronate-binding domain [44]. It remains to be seen whe the r this matrix is not permissive for axonal growth thus explain- ing the absence of axonal regenera t ion in adult mam- malian spinal cord.

Acknowledgements

We are grateful to Dr. Underhi l l for providing us the [3H]HA and Dr. Ashe r for the monoclonal anti- bodics. Wc would like to thank William S. Lane. Ruth J. Davenpor t and Renee A. Robinson of the Harva rd Microchemistry Facility, Cambr idge MA, for their ex- pert protein sequencing and amino acid analyses. Oligosaccharides from hyaluronic acid were p repa red by Dr. Filippo Biviano at Fidia Research Laborator ies , Abano T e r m e . Italy. This work was suppor ted by Uni ted States Public Hea l th Service Gran t NS 13034

and by the D e p a r t m e n t of Ve te rans Affairs.

References

1 Ilorstmann. E. and Meres, It. (19591 Z. Zellforsch. Microsk. Anat. 49. 569-604.

2 Van Harreveld. A.. Crowell. J. and Malholra. S.K. (19651 J. Cell Biol. 25. 5411-5415.

3 Nicholson. C. and Rice, M.E. (19861 Ann. NY Acad. Sci. 481, 55-71.

4 Perides. G.. Lane. W.S.. Andrews D., Dahl. D. and Bignami. A. (19891 J. BioL Chem. 282. 5981-5987.

5 Zimmermann. H. and Ruoslahti. E. (1989) EMBO J. 8. 2975- 2981.

6 Delpech. B. and Halvent, C. (19811 J. Neurochem. 36.855-859. 7 Delpech. B.. Mangonnat. C. Delpech. A.. Maes. P., Girard. N.

and Bertrand. P. (19911 Int. J. Biochem. 23, 329-337. 8 Caterson. B.. Baker. J.R.. Cristner. J.E.. Lee. Y. and Lentz. M.

(19851 J. Biol. Chem. 260, 11348-11356. 9 Caterson. B.. Calabro. T.. Donohle. P.J. and Jahnke. M.R. t1986)

in Articular Cartilage Biochemistry. (Schleyerback. R. and Has- call. V.C.. eds.), pp. 59-73. Raven Press, New York.

It) Underhill. C.B.. Chi-Rosso. G. and Topic. B.P. (1983) J. Biol. Chem. 258. 8086-8090.

I I Spiro, R.G. (1%6) Methods En~mol. 8, 3-26. 12 Davidson, E.A. (1966) Methods Enzym01.8, 52-60.

13 Dische. Z. and SheUles. L.B. (19691 J. Biol. Chem. 175. 595-603. 14 Park. J.T. and John.son M.J. (19491J. BioL Chem. 181. 149-151. 15 Warren. L. (1959) J. Biol. Chem. 234. 1971-1975. 16 Levvy. G.A. and McAllan. A. (19591 Biochem. J. 73. 127-132. 17 Elson. L.A. and Morgan. W.T.J. (19341Biochcm. J. 27.1824-1833. 18 Lowry. O.H.. Roserhrou~, N.J., Fort. A.L. and Randall. R.J.

(19511J. Biol. Chem. 193. 262-27. 19 Weissmann. B.. Meyer. K.. Sampson. P. and Linker. A. (19541 J.

Biol. Chem. 208. 417-429. 20 Laemmli. U.K. 0970) Nature 227. 680-685. 21 Cells. J.E. and Bravo. R. (1985) in Two-dimensional Gel Elee-

trophoresis of Proteins. Meth~s and Applications. pp. 3-3fl. Academic PressOrlando. FL.

22 Morissey. J.H. (19811 Anal. Bit,chem. 117. 307-310. 23 Towhin. H.. Staehelin. T. and Gordon. J. ( 10701 Proc. N~.tl. Aead.

Sci. U.S.A. 76. 4350-4354. 24 Underhill. C.B.. Tarone. (3. and Kausz, A.T. (1987) Connect.

Tiss. Res. 16. 225-235. 25 Matsudaira. P.. (1987)J. Biol. Chem. 262. 10035-11~38. 26 Tengblad. A. (1979) Biochim. Biophys. Acta 576. 281-289. 27 Fatz. L.L.. Reddi. A.H., Hascall. O.K.. Martin. D.. Pita. J.C. and

I-In,all. V.C. (1978)J. Biol. Chem. 254. 1375-1380. 28 Asher. R.A.. Pericles. G.. Vanderhaegen. J.-J. and Bignami. A.

(19911 J. Neurosci. Research 28, 410-421. 29 Baker. J.R. and Caterson. B. ( 19791J. Biol. Chem. 254. 2387-2393. 30 Neame. P.J.. Christner, J.E. and Baker. J.R. (1986) J. Biol. Chem.

261. 3519-3535. 31 Deak. F.. Kiss. I.. Sparks. K.;., Aggraves. W.S.. Hampikian, G.

and Goetinck. P.F. (1986) Pro¢. Natl. Acad. Sci. USA 83,3766- 3770.

32 Aebersold. R.H., Leavitt. J. Saavedra, R.A.. Hood. L.E. and Kent. S.B.tt. (19871 Proc. Natl. Acad. Sci. USA 84. 6970-6974.

33 Goctinck. P.F.. Stirpe, N.S.. Tsonis. P.A. and Carlone. D. (19871 J. Cell Biol. 105. 2403-2408.

34 Hardingham. T.E. and Muir. H. (1972) Biochim. Biophys. Aela 279, 401-405.

35 Hascall. V.C. and Sajdera, S.W. (19691 J. Biol. Chem. 244. 2384-2396.

36 Docge. K.. Sasaki. M.. Horigan, E.. Hassell. J.R. and Yamada. Y. t1987) J. Biol. Chem. 262. 17757-17767.

37 Perkins, S.J.. Nealis. A.S.. Dudhia, J. and Hardingham. T.E. (19891 J. Mol. Biol. 206. 737-748.

38 London. M. (19771 Methods Enzymol. 47. 145-149. 39 Hascall. V.C. and Heinergard. D.. (1973) J. Biol. Chem. 249.

4242-4249. 40 Fosang, A.J. and Hardingham, T.E. (19891 Biochem. J. 261.

801-809. 41 Goldstein, L.A., Zhou. D.F.H., Picker. L.J.. Minty. C.N. Bar-

gatze. R. F.. Dine. J.R. and Butcher. E.C. (19991 Cell 61. 1063- 1072.

42 Stamenkovic. I., Amiot. M.. Pesando. J.M. and Seed. B. t1989) Cell 56. 1057-1062.

¢3 Aruffo. A.. Slamenkovic. I., Melnick, M.. Underhill. C.B. and Seed. B. (1990) Cell 61, 1303-1313.

44 Perris. R. and Johan~.~on. S. (19901 Dev. Biol. 137. 1-12.

interaction of a brain extracellular matrix protein with hyaluronic acid

Documents