202
BlOCHEMlCAL SOClETY TRANSACTlONS
3. Doege, K., Fernandez, P., Hassell, J . R., Sasaki, M. & Yamada, Y.
(1986)J. Biol. ('hem. 261,8108-81 1 I
4. Sai, S., Tanaka. T.. Kosher, R. A. & 'I'anzer. M. L. (1986) /'roc.
N d . Acad. Sci. U.S.A. 83. 5081-5085
5 . Oldberg, A., Antonsson, P. & Heinegard, D. ( 1987) Biochem. J.
7. Tanaka, 'I-., Har-el, R. & Tanzer, M. L. (1988) J. Biol. C'hem.
263,1583 1 - I5835
8 . Rhodes, C., Doege. K.. Sasaki, M. & Yamada, Y.( 1 9 x 8 ) J. Hi()/.
C'hem. 263,6063-6067
243,255-255,
6. Kimura. T.. Mattei. M.-G., Stevens. J. W., Goldring. M. U..
Ninomiya. Y. & Olsen. B. R. ( 1Y X Y ) Eitr. J. Hiochern. I 7 9 , 7 1-78
Received I 7 October I Y 8 Y
Domains in cartilage proteoglycans: do they define function?
PETER J. NEAME
Shriners Hospitulfiir C'rippled Children, 12.502 N. Pitie Drive,
Tumpu, FL 33612, U.S.A.
Proteoglycans ( PGs) have, until recently, been poorly characterized at the amino acid sequence level. Since 1985, when
the first proteoglycan core protein structure was determined
1 11, structural information on these complex molecules has
become relatively plentiful. Proteoglycans are a good
example of protein construction through the assembly of
domains. It is tempting to assume that the individual domains
define discrete functions: this will be discussed here.
The major proteoglycans of cartilage are the large, aggregating chondroitin sulphate proteoglycans and the small,
dermatan/chondroitin sulphate PGs (PG-S1 and PG-S2).
The protein cores of these PGs have recently been
sequenced. The sequence of the large PG from rat chondrosarcoma was determined by Doege el ul. 121 and the
sequences of small PGs were determined by Krusius &
Ruoslahti (PG-S2 from human skin) [ 31, Day et al. (PG-S2
from bovine bone) j4J, Fisher et ul. (PG-SI from human
bone) [ S j and Nearne et ul. (PG-SI from bovine cartilage) 16).
All of these molecules have discrete domains which are
amenable to theoretical analysis of their structure and function. In some cases, it has been possible to analyse their function experimentally and examine thcir structure at a gross
level by rotary shadowing.
will interact in a similar way to the light and heavy chains o f
an IgG V-region. Experimental evidence for this dimer exists
[ 141. The regions which would correspond to the p-sheets
are particularly conserved in four o u t o f the seven potential
sheets. The conserved regions correspond to internal psheets or structures involved in the close approach o f two
V-regions.
Based on this model, the aggregate of LP and GI would
have a group of six loops, corresponding t o the hypervariablc
loops in IgG, at one end of the pseudo V-region and an oligosaccharide at the other. The PTRs would originate at the
same end of the aggregate as the oligosaccharide.
Circular dichroism (c.d.) spectra of LP (Fig. I ) did not,
however, show the characteristic positive ellipticity of psheet-containing proteins which can be seen at 240 nm. In
Curtiluge uggregutitigproteoglycuti
This proteoglycan forms large aggregates with hyaluronic
acid ( H A ) and in this form makes up much of the bulk of
cartilage due to the water associated with its glycosaminoglycan (GAG)chains. The first discrete domain found in this
PG was the HA-binding region (HABR) 171. Recently, other
domains have been identified [8].This discussion will be
restricted to the two domains at the N-terminal, referrcd to
as G1 (HABR) and G2. These structures have a striking
resemblance to link protein (LP), which is a 40-50 kDa
glycoprotein required for the formation of large stable aggregates of the PG with HA. Primary structures are available for
rat, chick and pig LP IY-lI]. GI is similar to the whole o f
link protein, while G2 is similar t o the C-terminal-half o f
either LP or G1. This region contains two loops, the 'proteoglycan tandem repeats' or PTRs.
The N-terminal of LP and G1 is a loop which falls into the
family of the immunoglobulin folds 112) and is rather less
conserved between LP and G1. The immunoglobulin fold
contains seven antiparallel p-sheets. Perkins er ul. [ 131 have
analysed the similarity between IgG V-regions and LP and
GI and found that there is evidence for p-sheets in this
region. It is therefore reasonable to assume that G1 and LP
Abbreviations used: PG, proteoglycan; HA. hyaluronic acid:
HABR, HA-binding region; LP, link protein; PTR, proteoglycan
tandem repeat: GAG, glycosaminoglycan: vWF, von Willebrand
factor.
I
220
I
240
Wavclcngth (nm)
I
260
Fig. 1. C ' d . spectrirtn ofL1'
LP (approximately 100 pg/rnl) in 0.4 M-guanidine hydrochloride, 50 mM-Tris/HCI, pH 7.5, was analysed with and
without H A (0.5 mgiml). The HA was added to the LP in 4
M-guanidine and the mixture dialyscd t o a final concentration
of 0.4 M-guanidine.
1900
203
PROTEOGLYCANS
fact, the c.d. spectrea looked more like a-hclix with a negative ellipticity from 240 to 2 13 nm, at which point the ellipticity started to movc back up before thc signal was lost at
2 10 nm. While the spectrum has not been deconvoluted. if
the assumption that the first loop is mostly P-sheet holds
true, then it can be inferred that the PTRs are mostly helical
and therefore may be dominating the spectrum. There was
no detectable effect o f adding HA. The pattern of homology
o f PTRs with the lymphocyte homing receptor, the Hermes
antigcn [ I S ] , tends t o reinforce the possibility of an a-helix.
Strongly conserved residucs arc found with thrce less
strongly conserved residues between them. a pattern very
characteristic o f amphipathic helices.
There arc two geometric requirements for binding the
G 1 -LP dimer to HA. ( 1 ) Each monomer must have a binding site one decasaccharide long I 16).( 2 ) The two monomer
binding sites must be juxtaposed axially with each other to
accommodate two adjacent decasaccharidcs of a linear
molecule (a total length o f (15 A). However, as a segment of
HA longer than the sum o f the two HA-binding sites (GI and
LP) is protected from digestion with chondroitinase ABC
[ 171. the sites may bc as much as 90 A apart t o give an
aggregate binding site o f 100-200 A.
The globular structure produccd by the postulated intcraction o f the pseudo V-regions of GI and LP would, by
analogy with IgG, be approximately 40 Ax 40 Ax 48 A and
could not account for the binding site. Part, at least, o f the
HA-binding site must therefore derive from the PTRs, in
contrast to what we have suggested previobsly 1181. Small
an@e neutron scattering for GI [ l o ] indicated that it has an
elliptical shape and therefore at least one of the PTRs would
extend away from the pseudo V-regions. There is experimental evidence. in the form o f limited proteolysis of LP
bound to HA, for the involvement of PTRs in the interaction
with HA [ 141. A model, taking into account the locations of
the N-linked oligosaccharides, the size o f the HA-binding
sites and the likely nature of the interaction between LP and
G I , is shown in Fig. 2. This is a very simple model and
ignores the probable interaction between the PTRs and the
N-terminal domain which would be expected in a globular
protein.
Recent evidence indicates that G2 (two PTR loops) docs
not bind to HA, G1 or LP [20]. This agrees with what is
observed by electron microscopy [S1. If G2 is capable of folding in the same way as the PTRs in G 1 o r LP, then the implication is that the PTRs require the rest of G1 ( o r LP) to fold
up correctly. Alternatively, the postulated placement of N linked oligosaccharides in G2 121. which is different from that
in G 1 and LP, may disrupt the structure to an extent which is
incompatible with binding to HA.
There is no explanation for the tendency of LP to form
very high-mass aggregates when not bound to HA or PG.
Synthetic peptides, designed to mimic the N-terminal loop or
a PTR, have not assisted our analysis of structure and function, beyond indicating that these domains are rather insoluble, which was already known from the primary
structure determination. However, we d o know that the
‘sides’ of a PTR, i.e. the regions between the 1st and 2nd
and the 3rd and 4th cysteines are flexible enough to form a
closed loop, a configuration which they will not usually form
in the intact molecule. It remains to be seen whether this
degree of flexibility exists in the native structure. One, at
least, of the PTRs in PG has an N-linked oligosaccharide. It is
likely that they both do. These sites must be on the solventaccessible side of the PTR.
Smull cardluge PGs
These PGs are part of a family of ubiquitous molecules
which have, so far, been found in two forms, PG-S1 and PGVOl. 18
N-linked
oligosccharides
40 A
Fig. 2. A model for the iiiteructioris within the IiA-LI’-GI
tertiury complex
The LP and G1 domains are interacting via their N-terminal
IgG-V-region-like domains. N-linked ohgosaccharides
attached to these act as a spacer separating the PTR domains
from each other and enabling these to interact with HA
through two similar, but non-contiguous, stretches of the
polysaccharide. This simplistic model does not, however,
take into account the directionality of the HA. If LP and G1
are bound to HA in the same way, then they should be
oriented in the same direction.
S2. Both PGs have GAG attached near the N-terminal; PGS2 having one GAG and PGI having two GAGS in most
cases. The major structural feature o f these molcules is a
series of nine leucine-rich repeats which are a motif found in
a variety of other molecules. The core proteins of these P G s
are probably globular structures based on rotary shadowing
P11,
Leucine-rich repeats are found in variety of other molecules. The exact function of the repeats is unknown. However,
most of these molecules also have analogues for the other
structures found in the small PGs, specifically the disulphidebonded loops. It seems reasonable to suppose that all these
molecules fold up in a similar way. It has been suggested that
these repeats form a P-sheet arrangement with the hydrophobic residues in internal positions and the external residues defining the specificity of its interactions [22]. In a
comparison of PG-S 1 and PG-S2, there are clear differences.
With our current knowledge, these differences are not very
informative: the repeats may form a hydrophobic core in the
PG and may not be very accessible to the solvent.
There are two small, disulphide-bond-containing loops in
the small PGs, which are also found in other molecules.
Apart from the position of the cysteines in the N-terminal
loop and, by implication, the disulphide bonds, there are no
strikingly conserved features, though these structures are
very conserved between the PGs. They may provide much of
the non-GAG-related functionality of the molecule.
There is good evidence that PG-S2 limits the size of collagen fibrils [23]. However, the similarity of the PGs to platelet GP-lb suggests another function. It has been found that
BIOCHEMICAL SOCIETY TRANSACTIONS
204
3. Krusius, T. & Ruoslahti, E. ( 1986) /'roc. N d . Acacl. Sci. U.S.A.
i.Collagen
Fig. 3. I'ostiiluted firtictioti of the small I'Gs
adherence to sirbetidotheliirm
it1
plutelet
The diagram shows a possible role for small PGs in the vWFmediated adherence of platelets. In this model, vWF can
bind to collagen and/or small PGs. While this model has not
been experimentally verified, it is based on the similarity
between GP- 1 b and the small PGs.
the von Willebrand factor ( vWF )-mediated platelet adherence to endothelium is not entirely due to collagen. Given
that GP-lb binds to vWF, it is possible t o propose an additional mode of platelet adhesion (Fig. 3). In this model, vWF
is bound to collagen and/or small PG while GP-lb binds t o
the other half of the vWF dimer via a similar interaction.
It should, in the next few years, be possible to test the
validity of these models and define some of the functions of
the PGs at the submolecular level. Eventually, tertiary structure will be defined and PGs can be included among the
ranks of well-characterized molecules.
Supported by N.I.H. grant A R 35322 and a grant from T h e
Shriners Hospital for Crippled Children.
I . Bourdon, M., Oldherg, A,. Piersbacher, M. & Ruoslahti. E.
( 1985) /'roc. N d . Accid. Sci. U.S.A. 82, I32 1 - I325
2. Doege. K.. Sasaki. M.. Horigan. E.. Hassell. J. R. & Yamada. Y.
( 1987)J. Hiol. C'hem. 262. 17757- I7767
83,7683-7687
4. Day, A. A.. McQuillan. C. I., Terminc, J. I). & Young. M. K.
( 1 987) Hiochcm. J. 248, 801 -805
5. Fisher, L. W., Termine. J . 0. & Young. M. F, ( I 989) J. Hiol.
('hem. 264,4571 -4576
6 . Neame, P. J., Choi. H. U. & Roscnherg. L. C. (1981)) ./. Hiol.
C'hem. 264,8653-866 I
7. Heinegird. I). & Hascall. V. C. ( 1974) J . Rial. ('hcwi. 249,
4250-4256
8. Paulsson, M., Miirgelin. M.. Wicdemann. H.. kxirdmorc-Gray.
M., Dunham, D., Hardingham, T., Heinegard, I),, Timpl, K. &
Engel, J. ( I 987) Hiochem. 1. 245, 763-772
9. Neame, P., Christncr. J . & Baker. J . ( IY86) J . Riol. ('hiwr. 261.
35 19-3535
10. Deik. F.,Kiss, I., Sparks, K. J., Argraves. W. S.,Hampikian, Ci.
& Goetinck, P. ( 1 986) /'roc. N o / / . Aciiel. . S c i ll..SA. 83.
3766-3770
1 I . Dudhia. J . & Hardingham. -1. E. ( I Y X Y ) J. Alol. Ifiol. 206.
74Y-753
12. Bonnet, F.. Pkrin. J.. Lorcnzo. F.. J o l l t : ~ . J . & JollCs. P. ( I 986)
Hiochim. Hiophyc-.Actit 813, 152- I55
13. Perkins. S. J.. Nealis. A . S.. Dudhia. J . & Hardinghnm. '1. E.
(1089)J. Mol. H i d . 206, 737-748
14. Pirin. J.. Bonnet. F.. Thurieau. C. & JollCs. P. ( I 987) J. Biol.
Chcm. 262, 13269- 13272
15. Goldstein. L. A.. Zhou. I>. F. H.. Pickcr. L. J.. Minty. C. N..
Bargatze, K. E, Ding, J. F. & Butcher, E. C . (1989) (211
(C'nmhriclgr, Muss.) 56, 1063- I072
16. Hascall. V. C. & Heinegard. I). ( I YX9) J. Riol. ('hcm. 249,
4242-4249
17. Faltz, L. L., Caputo. C. B.. Kimura. J. H., Schrode. J . & Hascall,
V. C. ( 1979)J. Hiol. (%em.254, I38 I - I387
I x. N a m e . P.. Christner. J. & Baker, J. ( 1087) .J. tfio/. ( ' h c w i . 262.
17768-17778
19. Perkins, S.. Miller, A,. Hardingham. T. & Muir. H. (1981) J.
Mol. Hiol. 150, 69-05
2 0 . Fosang. A . J . & Hardingham. T, E. ( I 981)) I f i o c h c w i . J. 261.
no I -XOY
21. Miirgelin, M., Paulsson, M., Malmstriim, A. & Heinegard, D.
( I 989) J. Hiol. ('hem. 264, 12080- I2090
22. Takahashi. N.. Takahashi. Y. & Putnam. F. W. (1985) /'roc
Neil/. A ( d . Sci. U..S.A. 82, I 906- I9 I 0
23. Vogel, K., Paulsson. M. & Heinegard, I). ( IU84) Rioc./rcvn. J.
223. 587-597
Received 1 7 October 1 Y89
Assembly of cartilage proteoglycan with hyaluronate and structure of the central filament in
proteoglycan aggregate
MATTHIAS MORGELIN, MATS PAULSSON and
JURGEN ENGEL
A bteihing Biophysikulische C'hemie, Biozetitrirrn, C'tl-4056
Basel, Switzerland
Proteoglycans are a family of large complex macromolecules
abundant in the extracellular matrix of all connective tissues,
but also present at cell surfaces. Each proteoglycan consists
of a protein core, which is substituted with the characteristic
glycosaminoglycan chains and often also with shorter oligosaccharide structures (for reviews, see [ l , 21). The major
proteoglycan in hyaline cartilage is a high-M, aggregating
species ( M , 1 x 10'-4 x 10") with a large extended core
protein to which typically about 100 chondroitin sulphate
chains, 30 keratan sulphate chains and about 6 0 short oligosaccharides are attached. The best understood interaction of
this proteoglycan is that with hyaluronate, which is an
unbranched polysaccharide present in cartilage. Aggregates
are formed by the non-covalent interaction between a
hyaluronate-binding region o f the protein core and a decasaccharide segment of the hyaluronate chain [3,41.The interaction is stabilized by the additional binding of link protein
which has affinity for both the binding region o f the
proteoglycan and the hyaluronate t o which it is attached,
such that a ternery complex is formed between these three
components. Many proteoglycans may bind to each hyaluronate chain and form aggregates with a length of as much as
4000 nm and a diameter of 500-600 nm (for reviews, see
[1,21).
Pioneering electron microscopic work by the KleinSchmidt technique [ 5 1 revealed binding of proteoglycan
monomers to the hyaluronate strands [ h ]and distribution of
side-chain substituents along the protein core 171. Protein
domains within the core of cartilage proteoglycan were first
resolved by rotary shadowing of proteoglycans sprayed on to
mica [S]. Three globular domains ( G I , G2 and G 3 ) and two
extended regions ( E l and E2) can be distinguished (19, 101;
Fig. I ). G I is located at the N-terminus and is separated from
1990